'Big Short' investor Michael Burry says Nvidia's memo was 'disappointing' — and he's betting against it and Palantir

Michael Burry publicly renewed his critique of Nvidia and broader AI firms, calling Nvidia's analyst memo a series of 'straw man' rebuttals and warning that next‑generation AI chips could become functionally obsolete between 2026–2028. He flagged aggressive extension of depreciation lives at hyperscalers that could precipitate future writedowns, disclosed he holds bearish put options on Nvidia and Palantir (Scion previously reported combined notional exposure of about $1.1bn at end‑September) and noted Nvidia shares have fallen roughly 14% from their Nov. 3 high.

Analysis

Market structure: Burry’s public put positions and depreciation thesis disproportionately pressure capital‑intensive AI hardware names (NVDA, PLTR) and their suppliers while benefitting software/cloud incumbents (MSFT) that can flex CAPEX timing. If hyperscalers pause refresh cycles, GPU order cadence can compress by 20–40% over 6–18 months, re‑rating spot GPU pricing and OEM revenue visibility. Options and implied volatility will stay elevated for NVDA/PLTR; tech credit spreads could widen 25–75bps if visible writedowns occur.

Risk assessment: Tail risks include abrupt mass writedowns at hyperscalers (plausible 2026–2028 cliff) and regulatory export or data‑use constraints that curtail AI deployment, each causing >30% EPS revisions for exposed players. Immediately (days) expect headline‑driven IV spikes; short‑term (weeks–months) expect guidance revisions and CAPEX pauses; long‑term (years) the silicon roadmap (performance/Watt) determines structural demand. Hidden dependencies: accounting policy shifts and inventory accounting can create sharp earnings cliffs even if end demand recovers.

Trade implications: Tactical trades favor asymmetric downside protection: small, concentrated long put positions on NVDA/PLTR (6–12 months, 20–30% OTM) or put spreads to limit cost; pair trade long MSFT (1–2% portfolio) vs short NVDA (1–2%) to express durable cloud demand versus hardware cyclicality. Rotate 3–5% allocation out of pure GPU hardware into software/SaaS with recurring revs (MSFT, NOW) and buy volatility into next two earnings/guidance windows.

Contrarian angles: The market may underappreciate NVDA’s software lock‑in (CUDA) and gross margin resiliency — a >25% drop could create durable buying opportunities rather than structural failure. Historical analogue: 2000s infrastructure shakeouts saw hardware losers and software winners; survivors consolidated share. Watch for short‑squeeze risk if put open interest is concentrated and for management accounting clarifications that could neutralize Burry’s thesis.

Key Decisions for Investors

• Establish a 1.5% portfolio notional bearish hedge on NVDA via 9–12 month puts 25% OTM (or equivalent 25/40% put spread) sized to cost no more than 0.3% of portfolio, act within 30 days or on NVDA IV >120% to cap premium.
• Initiate a 0.75–1.0% portfolio put spread on PLTR (6–9 month, 20–30% OTM buy / 40% OTM sell) to express downside while capping premium; escalate to 1.5% if PLTR gaps >15% post‑earnings.
• Implement a relative trade: long MSFT (1.5% weight) financed by short NVDA equity (1.5% weight) to capture cloud resilience vs hardware cyclicality; review after 60–90 days or on NVDA guidance change.
• Reduce pure AI/hardware exposure by 3–5% and reallocate to enterprise SaaS/cloud (MSFT, NOW) and defensive tech with recurring revenue; target next 45 days and rebalance if NVDA falls >25% from recent highs.
• Monitor proximate catalysts: require NVDA quarterly guidance or hyperscaler CAPEX commentary within 30–60 days and 10‑Q depreciation policy notes; if companies extend useful lives by >33% versus prior filing, increase downside hedges by 50%.