The Big Short’s Michael Burry Predicts a Possible 1987-Style Drop and is Betting Against AI Giants Nvidia and Palantir. Should You Follow?

Michael Burry reiterated short positions in Nvidia, Palantir, Micron Technology, and the iShares Semiconductor ETF, warning that the market rally could culminate in a 1987-style decline. He characterized AI-industry calls to slow model development as a potential cover for slowing growth, amid elevated AI infrastructure spending, inflation concerns, geopolitical turmoil, and a Shiller CAPE ratio surpassed only during the dot-com bubble. The article nevertheless maintains that long-term AI demand and earnings growth remain strong, while cautioning that near-term returns may be weaker and more volatile.

Analysis

The investable signal is not the prominent bear’s view; it is whether AI capex converts from infrastructure revenue into customer-level ROI before hyperscaler depreciation, power, and networking costs rise further. NVDA remains the highest-quality earnings vehicle because its revenue is tied to already-budgeted accelerator deployments, while PLTR is more exposed to a multiple reset if commercial contract expansion decelerates even without an outright revenue miss. MU has a different risk profile: AI memory demand can support mix and pricing, but its equity remains vulnerable to the broader memory-cycle inventory correction rather than simply an AI sentiment unwind.

Near term, public skepticism around AI safety and growth can increase dispersion and create a de-risking impulse in crowded, high-beta software and semiconductor baskets. A broad short of SOXX/SMH is structurally imprecise: it combines companies monetizing current demand with firms that have far less AI earnings exposure, raising borrow, timing, and squeeze risk if capex guidance holds. The more relevant 1-3 month catalysts are hyperscaler capex commentary, NVDA supply/demand and gross-margin guidance, and enterprise software bookings that demonstrate whether AI pilots are moving into recurring production spend.

Contrarianly, a slowdown in frontier-model training would not necessarily impair the semiconductor thesis proportionally; inference deployment can broaden unit demand and shift value toward memory, networking, power, and enterprise implementation. The greater six-to-18-month risk is that AI infrastructure spending becomes a zero-sum share battle among cloud platforms, compressing their returns on invested capital and ultimately reducing the pace of incremental orders. A durable bear case requires evidence of canceled data-center commitments, lengthening lead times, or material downward revisions to capex—not commentary from a high-profile investor.

Positioning should therefore favor quality and relative-value expressions over a market-crash hedge. Monitor implied volatility and concentration: a sharp index drawdown could temporarily pull NVDA lower regardless of fundamentals, but it would create a more attractive entry only if forward demand and customer capex plans remain intact.

Key Decisions for Investors

• Do not follow a headline-driven outright short in NVDA or SOXX. Establish a watch-level for adding NVDA on a 10-15% fundamental-risk-off pullback, contingent on no reduction in next-two-quarter revenue outlook or hyperscaler AI capex plans; target a 15-25% recovery over 6-12 months, with exit on a material demand-guidance reset.
• Implement a 3-6 month relative-value pair: long NVDA / short PLTR in approximately beta-neutral dollars. This expresses near-term infrastructure monetization versus valuation-sensitive application software; take profit if the relative spread moves 15-20% in favor, and stop if PLTR materially accelerates commercial remaining-deal-value or operating-margin guidance while NVDA margins compress.
• Treat MU as a tactical rather than core AI long. Add only after memory pricing and inventory data confirm continued DRAM/HBM tightness; use a 2-4 month horizon and reduce exposure if management signals conventional DRAM supply additions or weaker non-HBM pricing, which would overwhelm the AI mix benefit.
• For portfolio protection, prefer a limited-risk 2-3 month QQQ put spread over single-name AI shorts if volatility remains moderate. This hedges the valuation/liquidity shock risk while preserving participation in company-specific AI earnings upside; reassess after the next major hyperscaler earnings cycle.