Michael Burry Just Revealed His Next Big Short, and It's a Bet Against Nvidia, Micron, and AMD

Michael Burry disclosed he was short the iShares Semiconductor ETF (SOXX) as of June 30, arguing AI infrastructure spending is unsustainable. SOXX trades at a rich 74.3x P/E, and Burry points to signs of demand softening—UBS survey data suggests ~60% of businesses are curbing AI spending—raising the risk of a sharp valuation-driven correction in the fund.

Analysis

This is less a statement on near-term chip demand than a bet that AI capex is moving from a scarcity phase to a discipline phase. If that transition is real, the first thing to break is not revenues but multiples: semis are priced like durable secular growers, so even a modest deceleration can drive disproportionate downside in SOXX versus the broader Nasdaq. The ETF’s concentration means the weakest balance sheets and most cyclical end-markets will matter more than the headline basket suggests.

The next 1-3 months are about confirmation, not conviction. The key catalyst is whether hyperscalers and model builders defend 2026 spend plans; if they do, a crowded short in a momentum-heavy ETF can squeeze sharply on any one strong guide. Over 6-18 months, the more important risk is inventory and utilization normalization, which would usually hit memory, equipment, and second-tier networking first before fully reaching the megacaps.

The contrarian point the market may be missing is that AI spending may not collapse so much as reallocate from frontier training hardware to lower-intensity inference and custom silicon. That would be negative for index beta but not equally negative across constituents, creating a better relative-value short than an outright thesis that "AI is over." The thesis is falsified if the next two hyperscaler earnings cycles show continued upward capex revisions and no sign of lead-time or order cancellation deterioration.

Key Decisions for Investors

• Buy 3-6 month SOXX put spreads on strength, sized tactically; target a 10-15% basket drawdown if capex commentary softens, with defined premium risk.
• Pair long GOOGL vs short SOXX into the next hyperscaler earnings season: express the view that monetized software platforms can outlast hardware capex multiples.
• Prefer relative shorts in MU and MRVL over the full ETF if you want higher beta to any AI-spend air pocket; these names are more exposed to inventory/ordering inflections than NVDA or AVGO.
• Use any post-earnings rally in NVDA/AVGO/AMAT as a timing window to add hedges; invalidate the short if forward capex guidance is raised again by major cloud buyers.