WATCH → small-long at the low end of the fair range (≈$175–195), not at market. Because NVDA's own cheap ~20× forward multiple already prices a sharp deceleration the bullish narrative denies — so the quality isn't the debate, the durability is; and on cycle-positioning NVDA is the Intel-2000 "ballast", not the Lucent/Pets "story" tier — it would fall later and on a multiple de-rate, not first on insolvency, but 2000 proves real cash didn't spare even Intel an ~80% drawdown.
NVDA FY2023 → FY2025 (GAAP, audited). Source: peer_comparison.md.
6 AI/semis peers, NVDA highlighted. Source: valuations NVDIA.md (stockanalysis.com, 2026-06-24).
Value ranges by method, built on forward multiples (not peer-median trailing — avoids the bloated band). Dashed line = current $197.15. Source: investment_memo.md §5 / valuation_models NVDA.
Hyperscaler capex doesn't flatten; 88%+ DC revenue base has something to lean on.
No real bite from AMD Instinct or custom ASICs.
Whales keep buying and new ones appear (SpaceX/Google-type wins).
The 2000 "bandwidth glut" analogue — capex without monetization.
One whale slows → revenue dents; NVDA's own ~$110B financing book (67% of rev) is the Lucent strand.
DeepSeek already showed a $588.8B one-day drop is possible.
The demand engine. Re-evaluate if capex guides flatten for 2+ quarters.
Mixed today: B200 $6.11→$4.22/hr soft; H100/A100 reportedly up. Re-evaluate if the slide broadens.
Does ~64%-of-AR confirm and rise? Is the ~$110B financing book growing vs revenue?
88.3% data-center; 3 customers each ≥10% of revenue; Burry: top-3 ~64% of AR (not independently verified).
FY2025 OCF/NI fell to 0.88 — AR +$13.1B, inventory +$4.8B absorbed cash.
China 13.1% of FY2025 revenue under U.S. export restrictions; further tightening removes a market.
Burry: hyperscalers extending GPU life to 5–6 yrs overstate AI ROI (~$176B 2026–28, his estimate) — threatens the capex that is NVDA's revenue.