NVIDIA is the AI-infrastructure stress test: exponential demand language meets a supplier with real gross margin and a CUDA lock-in story. Trend seats map the S-curve; quant seats ask whether the print is a cycle peak dressed as a secular trend.
Wood/Laffont adoption curves vs. Dalio/Marks cycle placement. Red team will invert: what if capex pauses, custom ASICs bite, or the customer concentration is one cloud buyer’s budget?
Customer concentration, export controls, inventory at hyperscalers, and how much of “data center” revenue is one generation of GPU. Filings are dense; read the concentration footnote before cheering TAM.
Independent briefs, then clerk assembly · 3 isolated prompts + 1 assembly pass · filing facts from SEC EDGAR (cached 7 days) — not investment advice, no buy button
Preview seating. Paid desks are 16 / 29 / 48 / 61. $49 is Associate (29), not 61. Partners is the same 61 with more compute.
NVIDIA CORP · 10-K FY · FY2026 · ended 2026-01-25 · CIK 0001045810
Filings from SEC companyfacts (us-gaap). P/B, P/S from cached price × diluted shares vs filings. Not a live quote. Not a live quote, not advice, not a buy button. Missing numbers stay blank; P/E shows n/m when diluted EPS ≤ 0.
Same file, different rules. Each brief states a view and why. Briefs are written in isolation so one loud narrative cannot drag the room. The clerk reads them only after they exist. Never “you should buy.” There is no buy button.
Constructive — the work is worth continuing, not an order. Owner test on NVDA: P/E 44.6, P/B 34.03, ROE 76.3%, net income $120.1B.
Question: “Research NVDA as a US-listed name.”. Owner earnings lens on NVDA. Plain Midwest English, folksy metaphors, long-horizon. Hard rule: Circle of competence: if the business cannot be explained simply, pass. Numbers on file: P/E 44.6; P/B 34.03; ROE 76.3%; net income $120.1B; revenue YoY 65.5%. On file: 10-K FY FY2026; revenue $215.9B; revenue YoY 65.5%; net income $120.1B; ROE 76.3%; LT debt/equity 0.05; P/E 44.6; P/B 34.0; P/S 24.8; price $218.36 cached. A low multiple that needs a hero story is not a margin of safety. A rich multiple with durable ROE can still be a business — but then price is the risk. Underwrite frame: Worth further homework.
Constructive — the work is worth continuing, not an order. Classify NVDA before cheering growth. revenue YoY 65.5%. P/E 44.6.
Question: “Research NVDA as a US-listed name.”. Classify before multiple. Six-category company classification system.. revenue YoY 65.5%. P/E 44.6. revenue $215.9B. On file: 10-K FY FY2026; revenue $215.9B; revenue YoY 65.5%; net income $120.1B; ROE 76.3%; LT debt/equity 0.05; P/E 44.6; P/B 34.0; P/S 24.8; price $218.36 cached. Fast growth with no category is how you buy a story. Stalwart growth with a broken category is how you overstay. Underwrite frame: Worth further homework.
Skeptical — the burden of proof is on the bull case. Invert NVDA first. What has to be false for the bull to die?
Question: “Research NVDA as a US-listed name.”. Invert NVDA. Invert: how would this thesis get killed? List the kill-shots before the romance. On file: 10-K FY FY2026; revenue $215.9B; revenue YoY 65.5%; net income $120.1B; ROE 76.3%; LT debt/equity 0.05; P/E 44.6; P/B 34.0; P/S 24.8; price $218.36 cached. Name what must be true for the bull: earnings quality, leverage (0.05), and whether footnotes hide the cycle. If I cannot name the kill path, I am not doing the job. Underwrite frame: Pass for now.
Counted from isolated briefs. Ratios are filings or cached model input — not a live quote, not a rating, not a buy button.
| Metric | Value | Note |
|---|---|---|
| P/E | 44.6 | P/B, P/S from cached price × diluted shares vs filings. Not a live quote. |
| P/B | 34.03 | P/B, P/S from cached price × diluted shares vs filings. Not a live quote. |
| P/S | 24.79 | — |
| ROE | 76.3% | — |
| LT debt / equity | 0.05 | — |
| Revenue (FY) | $215.9B | — |
| Revenue YoY | +65.5% | — |
| Net income | $120.1B | — |
| Diluted EPS | $4.90 | — |
Cycle, timing, exit, and quant seats are dark on this plan. A three-seat sample cannot locate the cycle or size a position.
Assembly is a second isolated pass over the briefs. Masters did not read each other. Research simulation — not advice, not a buy button.
Analyst unlocks 16 independent seats (16 isolated prompts + 1 assembly pass).
See AnalystCounted from independent briefs only — not a vote, not a recommendation.
Trend research (Wood, Laffont, and the rest of the era desk) is locked on Observer.
Unlocks on Analyst (16). Empty seats stay empty — we do not ghost-fill them.
Compare plansCycle desk (Dalio, Marks, Templeton) is locked on Observer.
Unlocks on Analyst (16). Empty seats stay empty — we do not ghost-fill them.
Compare plansWarren BuffettNVDA: screened alone against Moat & Management Final Review.
Peter LynchNVDA: screened alone against Ten-Bagger Hunter.
Debate starts only after independent briefs exist. Red team still does not see a shared draft.
Charlie Munger — Debate chair · inversion: the best way to kill this thesis
Warren Buffett — Final review · 10-year hold test
Full transcript, risk list, and sizing stay on higher seats.
Unlock more seatsTiming desk (O’Neil and the tape seats) is locked on Observer.
Unlocks on Analyst (16). Empty seats stay empty — we do not ghost-fill them.
Compare plansExit desk (Livermore, Darvas, Bogle accountability) is locked on Observer.
Unlocks on Committee or Partners (61). Empty seats stay empty — we do not ghost-fill them.
Compare plansQuant (Thorp Kelly sizing, then the full seven on Committee) is locked on Observer.
Unlocks on Analyst (16). Empty seats stay empty — we do not ghost-fill them.
Compare plansAgents61 is a research simulation and educational tool. Isolated views, splits, and assembly notes only. Not personalized advice. AI-generated.
