{"version":"market-claims.v1","expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"stats":{"sources":26,"marketClaims":50,"explicitCalls":0,"directionalTheses":44,"ambiguousMentions":0,"openClaims":49,"median30dReturn":null,"lastUpdatedLabel":"recently"},"claims":[{"id":"mcl_a3820ed95823f9178d","url":"/trace/claims/mcl_a3820ed95823f9178d","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"TSMC runs the only factory AI can't do without, N3 demand is already at 110-120% of available capacity, but it hasn't raised prices to match the scarcity, a gap that closes in its favor.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"TSM","name":"TSM","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2077031014396756438","sourceUrl":"https://x.com/semianalysis_/status/2077031014396756438","timecode":null,"startMs":null,"endMs":null,"excerpt":"N3 utilization expected to exceed 100% in the second half of 2026"},"performance":{"entryPrice":424.61,"currentPrice":420.39,"returnPct":-0.993853,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_e3fa89cc110cfbaabe","url":"/trace/claims/mcl_e3fa89cc110cfbaabe","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Micron is about to get a massive repricing: DRAM fabs are running above 90% utilization with no relief in sight, and AI inference is the demand engine that won't stop growing.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"MU","name":"MU","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2077031014396756438","sourceUrl":"https://x.com/semianalysis_/status/2077031014396756438","timecode":null,"startMs":null,"endMs":null,"excerpt":"memory prices having gone up 6x in the past year"},"performance":{"entryPrice":968.51,"currentPrice":983.12,"returnPct":1.508503,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_ea326d35fe7eedc67e","url":"/trace/claims/mcl_ea326d35fe7eedc67e","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Storage just became the main event in AI infrastructure, with both NVIDIA Vera Rubin and AMD Helios now integrating in-rack storage for long-context AI workloads. Micron supplies the HBM4 memory in both platforms, faces a 25% DRAM supply deficit by 2030, and Goldman Sachs projects DRAM prices rising 250-300% in 2026.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"MU","name":"MU","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2077070734124281997","sourceUrl":"https://x.com/semianalysis_/status/2077070734124281997","timecode":null,"startMs":null,"endMs":null,"excerpt":"And why storage just became the main event."},"performance":{"entryPrice":985.9,"currentPrice":983.12,"returnPct":-0.281976,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_721baaa77da62d1617","url":"/trace/claims/mcl_721baaa77da62d1617","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Vera Rubin is now shipping to 8 cloud partners and more than 95% of the prior-gen Blackwell GPU capacity hasn't even been deployed yet, meaning the full NVIDIA rack-scale upgrade cycle is still in its earliest innings with Meta alone committing $125-145B in AI capex for 2026.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2077070734124281997","sourceUrl":"https://x.com/semianalysis_/status/2077070734124281997","timecode":null,"startMs":null,"endMs":null,"excerpt":"NVIDIA Vera Rubin. Double-wide racks. Liquid cooling. And why storage just became the main event."},"performance":{"entryPrice":209.84,"currentPrice":211.8,"returnPct":0.934045,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_23adec55c5f06d308c","url":"/trace/claims/mcl_23adec55c5f06d308c","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"AMD's first-ever rack-scale AI platform goes into production this half with OpenAI (6 GW) and Meta ($60B, 6 GW) already signed, and the CEO takes the stage July 23 to unveil MI450 specs — the biggest product moment in AMD's history against NVIDIA.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"AMD","name":"AMD","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2077070734124281997","sourceUrl":"https://x.com/semianalysis_/status/2077070734124281997","timecode":null,"startMs":null,"endMs":null,"excerpt":"AMD Helios. NVIDIA Vera Rubin. Double-wide racks. Liquid cooling. And why storage just became the main event."},"performance":{"entryPrice":554.64,"currentPrice":548.13,"returnPct":-1.173734,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_43041e5d7763653ca1","url":"/trace/claims/mcl_43041e5d7763653ca1","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Supermicro is the rack-and-stack integrator for every major AI platform shipping in 2026, and the CBO just went on record outlining why storage is the new battleground — both are revenue tailwinds for the only company first-to-market on Vera Rubin NVL72 and AMD Helios simultaneously.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"SMCI","name":"SMCI","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2077070734124281997","sourceUrl":"https://x.com/semianalysis_/status/2077070734124281997","timecode":null,"startMs":null,"endMs":null,"excerpt":"AMD Helios. NVIDIA Vera Rubin. Double-wide racks. Liquid cooling. And why storage just became the main event."},"performance":{"entryPrice":27.85,"currentPrice":27.65,"returnPct":-0.718133,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_7e35f4e0dbde9fa769","url":"/trace/claims/mcl_7e35f4e0dbde9fa769","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"NVIDIA wins by default as AI chip startup compiler promises fall apart. Rivals claim their hardware can run any model with zero manual tuning, but that's not how it works, and customers keep coming back to CUDA.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2076350972494807271","sourceUrl":"https://x.com/semianalysis_/status/2076350972494807271","timecode":null,"startMs":null,"endMs":null,"excerpt":"Laughing at all the AI chip startups that claim to have a magic compiler capable of mapping any model onto their chips efficiently, without any manual tuning."},"performance":{"entryPrice":212.65,"currentPrice":210.96,"returnPct":-0.794733,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_713e3bdbd57f8fb4bc","url":"/trace/claims/mcl_713e3bdbd57f8fb4bc","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"SHORT","claimText":"AMD's flagship AI GPU delivers less than half the token throughput of NVIDIA's B300 in the exact workloads hyperscalers are spending on most, making it hard to justify a premium AI valuation.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"AMD","name":"AMD","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075988793451069782","sourceUrl":"https://x.com/semianalysis_/status/2075988793451069782","timecode":null,"startMs":null,"endMs":null,"excerpt":"NVIDIA mogs AMD."},"performance":{"entryPrice":566.84,"currentPrice":557.89,"returnPct":1.578929,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_72f28b2f91d3e28fb4","url":"/trace/claims/mcl_72f28b2f91d3e28fb4","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"NVIDIA's newest GPUs outperform AMD's best AI chip by more than 2x on real-world LLM inference, the benchmark that matters most for cloud customers choosing hardware today.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075988793451069782","sourceUrl":"https://x.com/semianalysis_/status/2075988793451069782","timecode":null,"startMs":null,"endMs":null,"excerpt":"NVIDIA mogs AMD. Also great to see B300 faster than B200."},"performance":{"entryPrice":209.43,"currentPrice":210.96,"returnPct":0.730554,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_4a0290e5bb68642412","url":"/trace/claims/mcl_4a0290e5bb68642412","status":"PUBLISHED","claimType":"EVENT_PREDICTION","direction":"LONG","claimText":"US imports from Taiwan just flipped above China for the first time, hitting $24.6B vs China's $23.5B in May 2026, driven entirely by AI chip demand, and TSMC manufactures the vast majority of those chips.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"TSM","name":"TSM","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075928369099661642","sourceUrl":"https://x.com/semianalysis_/status/2075928369099661642","timecode":null,"startMs":null,"endMs":null,"excerpt":"We imported more from Taiwan than China as tariffs and AI trends collide."},"performance":{"entryPrice":438.46,"currentPrice":434.11,"returnPct":-0.992109,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_35a20cb1c140b470bf","url":"/trace/claims/mcl_35a20cb1c140b470bf","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"SHORT","claimText":"Apple's answer to losing the AI race is a lawsuit, not a model. Filing against OpenAI today signals the partnership is dead, Siri is still powered by Google's Gemini (not Apple's own AI), and Apple's hardware talent has walked out the door to build for rivals.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"AAPL","name":"AAPL","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075696326277222551","sourceUrl":"https://x.com/semianalysis_/status/2075696326277222551","timecode":null,"startMs":null,"endMs":null,"excerpt":"APPLE IS CATCHING UP IN AI."},"performance":{"entryPrice":314.96,"currentPrice":315.32,"returnPct":-0.1143,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_0673bb1bc19c08ab1c","url":"/trace/claims/mcl_0673bb1bc19c08ab1c","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Google committing to coherent-lite at scale for its AI chip interconnects turns a niche optical architecture into a hyperscaler volume product, lifting demand for coherent optical component makers already supply-constrained on EML lasers and transceivers.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"LITE","name":"LITE","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075566643573400052","sourceUrl":"https://x.com/semianalysis_/status/2075566643573400052","timecode":null,"startMs":null,"endMs":null,"excerpt":"we are finally expecting to see volume adoption at Google"},"performance":{"entryPrice":767.21,"currentPrice":771.54,"returnPct":0.564383,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_4fa31c3209d9c50f4c","url":"/trace/claims/mcl_4fa31c3209d9c50f4c","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Marvell lands the custom DSP inside Google's next-gen AI networking chips, a named design win at scale that extends its streak of hyperscaler custom silicon wins beyond AWS and Meta into Google's TPUv9 build-out.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"MRVL","name":"MRVL","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075566643573400052","sourceUrl":"https://x.com/semianalysis_/status/2075566643573400052","timecode":null,"startMs":null,"endMs":null,"excerpt":"Marvell working on the custom DSP"},"performance":{"entryPrice":237.85,"currentPrice":null,"returnPct":null,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_f8675c3c19833753d8","url":"/trace/claims/mcl_f8675c3c19833753d8","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Nvidia gets a revenue surge as Meta runs the most aggressive compute buildout ever documented, flooding money into GPU procurement to fuel its frontier AI push.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075377602647429413","sourceUrl":"https://x.com/semianalysis_/status/2075377602647429413","timecode":null,"startMs":null,"endMs":null,"excerpt":"the most aggressive compute ramp we've ever seen"},"performance":{"entryPrice":201.43,"currentPrice":202.78,"returnPct":0.670208,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_a0a20b5100be2824ec","url":"/trace/claims/mcl_a0a20b5100be2824ec","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Meta pulls ahead in the AI race, the only company with world-class data, talent, and compute all at once, giving it the best shot at catching OpenAI and Anthropic.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"META","name":"META","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075377602647429413","sourceUrl":"https://x.com/semianalysis_/status/2075377602647429413","timecode":null,"startMs":null,"endMs":null,"excerpt":"We believe Meta is the only hyperscaler/neolab on track to be world class at all three and therefore has the best chance at catching up with Anthropic/OpenAI"},"performance":{"entryPrice":640.14,"currentPrice":631.48,"returnPct":-1.352829,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_319cb6ac52514bd10a","url":"/trace/claims/mcl_319cb6ac52514bd10a","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Five simultaneous gigawatt-scale AI clusters going up at once means GPU orders unlike anything NVIDIA has ever seen, and more than 95% of announced Blackwell capacity still hasn't shipped, leaving years of backlog ahead.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075298749660725591","sourceUrl":"https://x.com/semianalysis_/status/2075298749660725591","timecode":null,"startMs":null,"endMs":null,"excerpt":"Never in the history of humanity have we ever seen a full 1GW campus under construction simultaneously—the closest was AWS building 800MW in Indiana for Project Rainier—yet Meta has two of them right now"},"performance":{"entryPrice":204.38,"currentPrice":202.78,"returnPct":-0.782855,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_b70d458d862c54175d","url":"/trace/claims/mcl_b70d458d862c54175d","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"SHORT","claimText":"Google's AI position is collapsing faster than the market realizes: its best model underperforms GPT and Claude in real-world use despite winning benchmarks, its Windsurf acquisition hasn't produced a real coding product, and Meta is now targeting Google's number three slot within six months.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"GOOGL","name":"GOOGL","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075298749660725591","sourceUrl":"https://x.com/semianalysis_/status/2075298749660725591","timecode":null,"startMs":null,"endMs":null,"excerpt":"Google had a brief moment in the spotlight with Gemini 3 Pro and Nano Banana, but they've since faded dramatically"},"performance":{"entryPrice":357.63,"currentPrice":358.89,"returnPct":-0.352319,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_aa8e42f04fa5e9ac93","url":"/trace/claims/mcl_aa8e42f04fa5e9ac93","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Meta is on pace to own more AI compute than OpenAI and Anthropic combined by year-end, running the most aggressive datacenter ramp in human history, while its ad revenue funds it all without diluting shareholders.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"META","name":"META","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075298749660725591","sourceUrl":"https://x.com/semianalysis_/status/2075298749660725591","timecode":null,"startMs":null,"endMs":null,"excerpt":"the most aggressive compute ramp we've ever seen"},"performance":{"entryPrice":619.26,"currentPrice":631.48,"returnPct":1.973323,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_ed549a434b2a9c3db6","url":"/trace/claims/mcl_ed549a434b2a9c3db6","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"NVIDIA's next-gen AI rack is moving from blueprint to mass production, with the Vera Rubin NVL72's 72-GPU, 440kW rack entering its First At-Scale Production Token milestone and locking in the next wave of hyperscaler orders.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075277408047903190","sourceUrl":"https://x.com/semianalysis_/status/2075277408047903190","timecode":null,"startMs":null,"endMs":null,"excerpt":"Vera Rubin NVL72 is a beast, thanks to its extreme co-designed system architecture, and First At-Scale Production Token (FASPT) is fast approaching."},"performance":{"entryPrice":203.5,"currentPrice":204.02,"returnPct":0.255528,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_fb80d557a4c0492a12","url":"/trace/claims/mcl_fb80d557a4c0492a12","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"NVIDIA's next-gen Rubin rack is nearly in production, packing 72 GPUs and 40 proprietary NVSwitch6 chips per system, with first at-scale token delivery approaching fast.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/bubbleboi/status/2075288558555082993","sourceUrl":"https://x.com/bubbleboi/status/2075288558555082993","timecode":null,"startMs":null,"endMs":null,"excerpt":"Vera Rubin NVL72 is a beast, thanks to its extreme co-designed system architecture, and First At-Scale Production Token (FASPT) is fast approaching."},"performance":{"entryPrice":204.27,"currentPrice":204.02,"returnPct":-0.122387,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_f647ccde6a18545926","url":"/trace/claims/mcl_f647ccde6a18545926","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Broadcom is cashing in on every TPU chip ordered, with Anthropic's nearly 1 million TPUv7 units, Meta's Iris ASIC, and OpenAI's custom inference chip all running through Broadcom's fabs, making it the quiet winner of the anti-Nvidia trade.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"AVGO","name":"AVGO","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075203875062120785","sourceUrl":"https://x.com/semianalysis_/status/2075203875062120785","timecode":null,"startMs":null,"endMs":null,"excerpt":"Google TPUs continue to gain market share outside of internal Google workloads, giving frontier AI labs a viable alternative for training."},"performance":{"entryPrice":402.95,"currentPrice":401.55,"returnPct":-0.347438,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_2f49543d75f326293a","url":"/trace/claims/mcl_2f49543d75f326293a","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"SHORT","claimText":"Nvidia's pricing power rests on being the only viable option for AI training, but Anthropic just proved that isn't true anymore. Five generations of frontier Claude models were trained on Google's chips, not Nvidia's, at roughly half the cost. An expanding list of labs is said to be lining up for TPUs, which chips away at the monopoly premium that justifies Nvidia's valuation.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075203875062120785","sourceUrl":"https://x.com/semianalysis_/status/2075203875062120785","timecode":null,"startMs":null,"endMs":null,"excerpt":"Google TPUs continue to gain market share outside of internal Google workloads, giving frontier AI labs a viable alternative for training."},"performance":{"entryPrice":205.66,"currentPrice":201.57,"returnPct":1.988719,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_19161d5103a3af1c75","url":"/trace/claims/mcl_19161d5103a3af1c75","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Google's AI chip business is breaking out beyond its own walls, with Anthropic's five successful Claude generations validating TPUs as a real alternative to Nvidia for frontier model training. A lineup of new external customers is forming behind Anthropic, which means Google Cloud earns revenue from every chip rented or sold.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"GOOGL","name":"GOOGL","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075203875062120785","sourceUrl":"https://x.com/semianalysis_/status/2075203875062120785","timecode":null,"startMs":null,"endMs":null,"excerpt":"Successfully training models on TPUs has been demonstrated by Anthropic through the past five-plus successful Claude releases."},"performance":{"entryPrice":357.06,"currentPrice":353.37,"returnPct":-1.03344,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_bca8eb83a761b0d61f","url":"/trace/claims/mcl_bca8eb83a761b0d61f","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Investors buying SpaceX for rockets are getting a $80B+ committed-revenue GPU cloud instead. Colossus is now rented to Anthropic at $1.25B/month, Google at $920M/month, and Reflection at $150M/month, turning idle AI compute into a recurring revenue engine that the market is only beginning to price.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"SPCX","name":"SPCX","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075022271563845638","sourceUrl":"https://x.com/semianalysis_/status/2075022271563845638","timecode":null,"startMs":null,"endMs":null,"excerpt":"With SpaceX finally going public, investors expected Space, Mars, and Skyward Rockets. Instead, they might see a galaxy of server racks lighting up."},"performance":{"entryPrice":149.95,"currentPrice":148.3,"returnPct":-1.100367,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_15742b5e36fdf9f491","url":"/trace/claims/mcl_15742b5e36fdf9f491","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"SpaceX's IPO isn't a rockets story — it's a GPU story, and every GPU rack in Colossus is an Nvidia chip sale. xAI is one of the biggest GB300 buyers this year, and the Colossus data center already has over 1 million Nvidia GPUs across sites, with more on the way.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2075022271563845638","sourceUrl":"https://x.com/semianalysis_/status/2075022271563845638","timecode":null,"startMs":null,"endMs":null,"excerpt":"As a major GB300 buyer this year, xAI is turning the space race into a GPU race."},"performance":{"entryPrice":203.72,"currentPrice":204.12,"returnPct":0.196348,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_04f97bd690d23f12c3","url":"/trace/claims/mcl_04f97bd690d23f12c3","status":"PUBLISHED","claimType":"EVENT_PREDICTION","direction":"LONG","claimText":"Google's $135B Anthropic stake is about to get a public price tag: Anthropic just turned profitable and is heading for an October Nasdaq listing at a $1T+ valuation, forcing the market to finally count what Google already owns.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"GOOGL","name":"GOOGL","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074803351728869865","sourceUrl":"https://x.com/semianalysis_/status/2074803351728869865","timecode":null,"startMs":null,"endMs":null,"excerpt":"Anthropic 3Q26 Profit Over $1B: The Anthropic IPO Financials Sneak Peak"},"performance":{"entryPrice":362.49,"currentPrice":367.03,"returnPct":1.252448,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_05978fb0d1f4f225bd","url":"/trace/claims/mcl_05978fb0d1f4f225bd","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Anthropic's token volume is set to grow from 5 quadrillion to 13 quadrillion per month by December alone, and that is just one lab. BofA projects Google, Meta, and Amazon collectively spending $816B on compute by 2028, a new street high, meaning demand for Nvidia GPUs is expanding faster than the market has priced.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074803351728869865","sourceUrl":"https://x.com/semianalysis_/status/2074803351728869865","timecode":null,"startMs":null,"endMs":null,"excerpt":"fund the massive AI buildout still to come"},"performance":{"entryPrice":193.83,"currentPrice":196.93,"returnPct":1.59934,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_8d3dd4ba44af3b05a4","url":"/trace/claims/mcl_8d3dd4ba44af3b05a4","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Anthropic hitting $1B quarterly profit and filing for IPO at a near-trillion-dollar valuation turns Amazon's massive bet into a winner. Amazon has up to $25B committed to Anthropic and is raising $200B this year for AI infrastructure, making AWS the financial backbone of the most profitable AI lab in the world.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"AMZN","name":"AMZN","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074803351728869865","sourceUrl":"https://x.com/semianalysis_/status/2074803351728869865","timecode":null,"startMs":null,"endMs":null,"excerpt":"Anthropic and OpenAI combine for ~$100B of ARR"},"performance":{"entryPrice":242.09,"currentPrice":245.98,"returnPct":1.60684,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_db2c1b1a477d7b5cb7","url":"/trace/claims/mcl_db2c1b1a477d7b5cb7","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Meta just raised its AI spending guidance to $125-145B for 2026, its highest ever, as the company builds out its own infrastructure rather than relying on third-party clouds, putting it squarely in the hyperscaler winner category.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"META","name":"META","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074659884398739889","sourceUrl":"https://x.com/semianalysis_/status/2074659884398739889","timecode":null,"startMs":null,"endMs":null,"excerpt":"AI spend continues to accelerate and hyperscalers are benefitting"},"performance":{"entryPrice":608.95,"currentPrice":615.58,"returnPct":1.088759,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_fd78056f68d9717e12","url":"/trace/claims/mcl_fd78056f68d9717e12","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Google Cloud is literally running out of capacity because AI demand is so intense, and Alphabet is doubling its capex from $90B to up to $190B in 2026 to catch up. When supply meets demand, GCP revenue accelerates.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"GOOGL","name":"GOOGL","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074659884398739889","sourceUrl":"https://x.com/semianalysis_/status/2074659884398739889","timecode":null,"startMs":null,"endMs":null,"excerpt":"AI spend continues to accelerate and hyperscalers are benefitting"},"performance":{"entryPrice":367.11,"currentPrice":367.03,"returnPct":-0.021792,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_bac31c79344bc3f205","url":"/trace/claims/mcl_bac31c79344bc3f205","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"AWS is pulling away as AI cloud spend surges, while semi stocks lag. Amazon is raising $25B in fresh bonds to pour into AI infrastructure and is expected to spend nearly $200B this year alone.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"AMZN","name":"AMZN","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074659884398739889","sourceUrl":"https://x.com/semianalysis_/status/2074659884398739889","timecode":null,"startMs":null,"endMs":null,"excerpt":"AI spend continues to accelerate and hyperscalers are benefitting"},"performance":{"entryPrice":245.69,"currentPrice":245.98,"returnPct":0.118035,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_e8b1c2eb4118fc0936","url":"/trace/claims/mcl_e8b1c2eb4118fc0936","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"NVIDIA's newest GPU just lapped its own previous generation by 3.7x on the metric that matters most to AI builders, making it the only real choice for inference at scale.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074599515256918065","sourceUrl":"https://x.com/semianalysis_/status/2074599515256918065","timecode":null,"startMs":null,"endMs":null,"excerpt":"delivering up to 3.7x faster interactivity than B300"},"performance":{"entryPrice":196.65,"currentPrice":196.93,"returnPct":0.142385,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_6d10a48d620579e7d5","url":"/trace/claims/mcl_6d10a48d620579e7d5","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"TSMC's real moat is a 93,000-item Silicon IP library that took 15 years to build, meaning any rival foundry must replicate decades of co-optimized design tools and IP before a single customer can switch.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"TSM","name":"TSM","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074539404454994333","sourceUrl":"https://x.com/semianalysis_/status/2074539404454994333","timecode":null,"startMs":null,"endMs":null,"excerpt":"TSMC's moat is bigger than PPA, EUV, or yield. It is the EDA/IP ecosystem wrapped around the fab."},"performance":{"entryPrice":436.36,"currentPrice":433.56,"returnPct":-0.641672,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_4ba59b8afb981c52e4","url":"/trace/claims/mcl_4ba59b8afb981c52e4","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Applied Materials wins twice as NAND layers multiply: its deposition tools are needed for every new layer, and it owns 9% of Besi, the company confirmed as the fastest hybrid bonding equipment maker in the world.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"AMAT","name":"AMAT","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074479087268045192","sourceUrl":"https://x.com/semianalysis_/status/2074479087268045192","timecode":null,"startMs":null,"endMs":null,"excerpt":"Both Kioxia and Samsung unveiled multi-array hybrid bonded NAND architectures at VLSI 2026 as a path to 1000+ layers for maximum density."},"performance":{"entryPrice":555.49,"currentPrice":528.85,"returnPct":-4.795766,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_685851720ba7f20571","url":"/trace/claims/mcl_685851720ba7f20571","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Every extra NAND layer means more tool passes, and Lam Research owns the top market share in etch and deposition, the two processes that multiply fastest as NAND stacks race past 1000 layers.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"LRCX","name":"LRCX","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074479087268045192","sourceUrl":"https://x.com/semianalysis_/status/2074479087268045192","timecode":null,"startMs":null,"endMs":null,"excerpt":"Both Kioxia and Samsung unveiled multi-array hybrid bonded NAND architectures at VLSI 2026 as a path to 1000+ layers for maximum density."},"performance":{"entryPrice":327,"currentPrice":314.22,"returnPct":-3.908257,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_41efae0508f8989ae4","url":"/trace/claims/mcl_41efae0508f8989ae4","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Kioxia has sold out every NAND chip it can make in 2026, and just unveiled a new stacking architecture that doubles the layers per chip, with flash prices up over 70% this quarter alone.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"KIOXIA","name":"KIOXIA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074479087268045192","sourceUrl":"https://x.com/semianalysis_/status/2074479087268045192","timecode":null,"startMs":null,"endMs":null,"excerpt":"Kioxia's Multi-Stacked cell Array (MSA) samples include a dual-218L (2-deck) stack to address mechanical integration challenges from alignment and warpage, and a dual-17L stack for electrical characterization and QLC reliability."},"performance":{"entryPrice":443.43,"currentPrice":null,"returnPct":null,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_da851f012069d048dd","url":"/trace/claims/mcl_da851f012069d048dd","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Samsung just built the world's densest NAND chip, a 900-layer class flash memory using a new bonding technique, while the entire NAND market is sold out and prices are surging 70-75%.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"SMSN","name":"SMSN","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074479087268045192","sourceUrl":"https://x.com/semianalysis_/status/2074479087268045192","timecode":null,"startMs":null,"endMs":null,"excerpt":"Samsung's Cell Multi-Bonding (CMB) samples went further, with a dual-450L (3-deck) mechanical sample and a dual-155L (1-deck) electrical sample."},"performance":{"entryPrice":192.57,"currentPrice":null,"returnPct":null,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_467b96523913e8bfd2","url":"/trace/claims/mcl_467b96523913e8bfd2","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"CoreWeave gets a six-year NVIDIA revenue guarantee on its GPU clusters, solving the financing chicken-and-egg problem that's kept Neoclouds from scaling, just as NVDA closes multiple gigawatt-scale datacenter deals in coming weeks that flow directly through operators like CoreWeave.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"CRWV","name":"CRWV","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/rhouseresearch/status/2074480399997182453","sourceUrl":"https://x.com/rhouseresearch/status/2074480399997182453","timecode":null,"startMs":null,"endMs":null,"excerpt":"NVIDIA has also been providing support by directly leasing capacity from colocation providers and then subleasing to Neoclouds"},"performance":{"entryPrice":84.342,"currentPrice":79.82,"returnPct":-5.361504,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_21164ae250830dc27f","url":"/trace/claims/mcl_21164ae250830dc27f","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Street is missing $33B in NVIDIA datacenter revenue for the back half of fiscal 2027, according to supply-chain tracking that shows NVDA acting as the central bank of AI, backstopping GPU rentals and leasing multiple gigawatts of colocation directly.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/rhouseresearch/status/2074480399997182453","sourceUrl":"https://x.com/rhouseresearch/status/2074480399997182453","timecode":null,"startMs":null,"endMs":null,"excerpt":"the general market is still materially lower on shipment volumes and revenue estimates for Nvidia in the 2nd half of this year versus our through supply chain tracking"},"performance":{"entryPrice":192.45,"currentPrice":191.81,"returnPct":-0.332554,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_35d3c765c8f5b9df3e","url":"/trace/claims/mcl_35d3c765c8f5b9df3e","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"The AI financing wave will be bigger than every US credit market except mortgages, and Nebius has pre-sold $46B worth of that buildout to Meta and Microsoft, with revenue already up 684% year-on-year to prove the model works.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NBIS","name":"NBIS","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074251380370423973","sourceUrl":"https://x.com/semianalysis_/status/2074251380370423973","timecode":null,"startMs":null,"endMs":null,"excerpt":"AI Debt Financing will become a multi-trillion-dollar credit market, with over $7T of debt outstanding by 2029 driven both by AI IT Capex and AI Datacenter Capex needs"},"performance":{"entryPrice":213.15,"currentPrice":213.02,"returnPct":-0.06099,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_e3feb2287deeb12b8e","url":"/trace/claims/mcl_e3feb2287deeb12b8e","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"CoreWeave and Neoclouds like it can now stand up GPU clusters without a hyperscaler offtake, because Nvidia itself backstops the revenue floor, making lenders willing to fund and driving a 25.4% IRR for 1-year rental books.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"CRWV","name":"CRWV","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074251380370423973","sourceUrl":"https://x.com/semianalysis_/status/2074251380370423973","timecode":null,"startMs":null,"endMs":null,"excerpt":"In the backstop program, Nvidia provides a take-or-pay commitment to Neoclouds – a minimum revenue guarantee on the underlying GPU capacity"},"performance":{"entryPrice":87.07,"currentPrice":86.46,"returnPct":-0.700586,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_9db09d4f81af27776a","url":"/trace/claims/mcl_9db09d4f81af27776a","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Wall Street is missing $33B of Nvidia datacenter revenue for 2H FY27, and a new backstop program embeds Nvidia into every Neocloud deal as the central bank of AI compute, taking a 40-60% revenue share above the floor.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074251380370423973","sourceUrl":"https://x.com/semianalysis_/status/2074251380370423973","timecode":null,"startMs":null,"endMs":null,"excerpt":"Nvidia revenue is skyrocketing, and even 3 years into the build out, the general market is still materially lower on shipment volumes and revenue estimates for Nvidia in the 2nd half of this year versus our through supply chain tracking"},"performance":{"entryPrice":195.64,"currentPrice":195.55,"returnPct":-0.046003,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_569bd3f1495434641e","url":"/trace/claims/mcl_569bd3f1495434641e","status":"PUBLISHED","claimType":"EVENT_PREDICTION","direction":"LONG","claimText":"SK Hynix lands on Nasdaq July 10 in the biggest ADR deal ever, finally giving US investors direct access to the dominant supplier of high-bandwidth memory powering every major AI chip.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"SKHX","name":"SKHX","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2074237106755359224","sourceUrl":"https://x.com/semianalysis_/status/2074237106755359224","timecode":null,"startMs":null,"endMs":null,"excerpt":"SKHY set to Sky-rocket 🚀? Is the name itself literally a spoiler?\nCountdown to the Nasdaq listing on July 10 starts now!"},"performance":{"entryPrice":1551,"currentPrice":null,"returnPct":null,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_9ee1e9a50545a502f3","url":"/trace/claims/mcl_9ee1e9a50545a502f3","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"Copper racks dominate AI infrastructure spending through 2027 with no next-gen competition: Kyber pushed to 2028, replacement cancelled, leaving Rubin Ultra as the only game in town for hyperscalers who can't wait.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/zephyr_z9/status/2073976008848601475","sourceUrl":"https://x.com/zephyr_z9/status/2073976008848601475","timecode":null,"startMs":null,"endMs":null,"excerpt":"Copper gets a big W"},"performance":{"entryPrice":195.15,"currentPrice":194.83,"returnPct":-0.163976,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_7094f0d00f29f2ce9b","url":"/trace/claims/mcl_7094f0d00f29f2ce9b","status":"PUBLISHED","claimType":"EVENT_PREDICTION","direction":"SHORT","claimText":"Two NVIDIA next-gen rack products just got knocked out, Kyber delayed to 2028 and NVL72x2 cancelled, stranding Rubin Ultra with limited scale-up options and a narrower competitive moat than investors priced in.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/au_xbt/status/2073928940876710296","sourceUrl":"https://x.com/au_xbt/status/2073928940876710296","timecode":null,"startMs":null,"endMs":null,"excerpt":"MASSIVE DELAY: Just 3 months after Jensen demoed Kyber NVL144 at GTC, it has faced major setbacks and has been delayed by more than 12 months, pushing it back to 2028."},"performance":{"entryPrice":195.63,"currentPrice":194.83,"returnPct":0.408935,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_264f62df061505e97d","url":"/trace/claims/mcl_264f62df061505e97d","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"LONG","claimText":"AMD stands to win datacenter GPU business that NVIDIA can't deliver for over a year, as Kyber is delayed to 2028 and the backup rack architecture was cancelled outright, leaving enterprise buyers looking for alternatives.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"AMD","name":"AMD","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/damnang2/status/2073900457215602756","sourceUrl":"https://x.com/damnang2/status/2073900457215602756","timecode":null,"startMs":null,"endMs":null,"excerpt":"MASSIVE DELAY: Just 3 months after Jensen demoed Kyber NVL144 at GTC, it has faced major setbacks and has been delayed by more than 12 months, pushing it back to 2028."},"performance":{"entryPrice":539.07,"currentPrice":517.82,"returnPct":-3.941974,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_fb247f9162a3b14f54","url":"/trace/claims/mcl_fb247f9162a3b14f54","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"SHORT","claimText":"Nvidia's next-gen server rack is pushed to 2028 and its backup architecture cancelled, just three months after Jensen demoed it on stage — a pattern of overpromising that leaves the company with nothing to ship.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/bubbleboi/status/2073901348534563062","sourceUrl":"https://x.com/bubbleboi/status/2073901348534563062","timecode":null,"startMs":null,"endMs":null,"excerpt":"Nvidia's motto is over promise and under deliver."},"performance":{"entryPrice":198.15,"currentPrice":194.83,"returnPct":1.675498,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_6de422b0545d4045aa","url":"/trace/claims/mcl_6de422b0545d4045aa","status":"PUBLISHED","claimType":"EVENT_PREDICTION","direction":"SHORT","claimText":"Lumentum's CPO optical revenue gets pushed out at least a year as NVIDIA's Kyber rack slips to 2028 and its NVL72x2 replacement is cancelled, removing the key platform that was supposed to drive co-packaged optics shipments this cycle.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"LITE","name":"LITE","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/damnang2/status/2073900457215602756","sourceUrl":"https://x.com/damnang2/status/2073900457215602756","timecode":null,"startMs":null,"endMs":null,"excerpt":"MASSIVE DELAY: Just 3 months after Jensen demoed Kyber NVL144 at GTC, it has faced major setbacks and has been delayed by more than 12 months, pushing it back to 2028."},"performance":{"entryPrice":757.32,"currentPrice":728.32,"returnPct":3.829293,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_93ac853d916221ea02","url":"/trace/claims/mcl_93ac853d916221ea02","status":"PUBLISHED","claimType":"EVENT_PREDICTION","direction":"SHORT","claimText":"NVIDIA's flagship next-gen rack system slips over a year to 2028, just three months after Jensen Huang personally demoed it on stage, and the back-to-back rack replacement was quietly cancelled, leaving Rubin Ultra with a limited scale-up ceiling.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/damnang2/status/2073900457215602756","sourceUrl":"https://x.com/damnang2/status/2073900457215602756","timecode":null,"startMs":null,"endMs":null,"excerpt":"MASSIVE DELAY: Just 3 months after Jensen demoed Kyber NVL144 at GTC, it has faced major setbacks and has been delayed by more than 12 months, pushing it back to 2028."},"performance":{"entryPrice":198.17,"currentPrice":194.83,"returnPct":1.685422,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}},{"id":"mcl_79c31c8891bdd2893e","url":"/trace/claims/mcl_79c31c8891bdd2893e","status":"PUBLISHED","claimType":"DIRECTIONAL_THESIS","direction":"SHORT","claimText":"NVIDIA's next-gen AI rack slips over a year to 2028, and the backup plan is scrapped entirely, leaving a gaping hole in its highest-end product roadmap just months after it was publicly demoed.","confidence":0.72,"expert":{"slug":"semianalysis","displayName":"SemiAnalysis","kind":"PERSON","status":"UNCLAIMED","avatarUrl":null,"url":"/trace/experts/semianalysis","claimed":false},"primaryEntity":{"kind":"ASSET","assetClass":"EQUITY","symbol":"NVDA","name":"NVDA","exchange":null,"confidence":0.78},"evidence":{"evidenceUrl":"https://x.com/semianalysis_/status/2073874671498387899","sourceUrl":"https://x.com/semianalysis_/status/2073874671498387899","timecode":null,"startMs":null,"endMs":null,"excerpt":"Just 3 months after Jensen demoed Kyber NVL144 at GTC, it has faced major setbacks and has been delayed by more than 12 months, pushing it back to 2028."},"performance":{"entryPrice":197.97,"currentPrice":194.83,"returnPct":1.586099,"relativePct":null,"method":"PASTE_AUTHOR_PRICE"}}]}