Michael Burry Is Staying Short Nvidia, Palantir, and Tesla, Warning of a Possible "1987-Type Fall" for AI Stocks. Should Investors Take the Bet Seriously?

Margin debt rose 50% year over year in June, heightening the risk that forced selling and margin calls could amplify a correction in AI and other growth stocks. The article argues Michael Burry's shorts in Nvidia, Palantir and Tesla overlook solid AI demand, revenue growth and improving profitability, though leveraged positioning remains a market-wide vulnerability. It notes Burry's disclosed Nebius short on Aug. 6 was followed by roughly a 20% gain in the stock after strong earnings.

Analysis

The relevant risk is not a broad AI-demand collapse but procyclical deleveraging in the most crowded, high-beta expressions. Daily-reset leveraged products mechanically amplify both directions, while margin liquidation transmits through liquid index constituents first; NVDA and TSLA can therefore decline materially even if their earnings estimates are unchanged. In a 1-4 week risk-off episode, semiconductor ETFs (SOXX/SMH) and momentum baskets are likely to be better liquidity shorts than single-name fundamental shorts.

Within AI, dispersion should widen rather than collapse uniformly. NVDA's valuation remains most sensitive to hyperscaler capex guidance and gross-margin durability, whereas PLTR is more exposed to multiple compression because a larger share of its valuation depends on sustained commercial-growth upside. NBIS has higher financing, customer-concentration, and execution sensitivity than the established platform names; a post-results rally does not validate its long-duration cash-flow assumptions.

The contrarian point is that public disclosures of bearish positions do not establish current exposure, strike, maturity, or hedge structure; they are weak evidence for a durable directional thesis. A forced-flow drawdown becomes investable only if it produces a disconnect between prices and forward estimates. Falsification for the constructive AI view would be consecutive hyperscaler capex cuts, NVDA data-center revenue deceleration accompanied by gross-margin erosion, or a persistent rise in credit spreads rather than a short-lived equity-volatility spike.

Key Decisions for Investors

• Maintain core NVDA exposure but do not add into a volatility spike; add only after a 10-15% drawdown if next-twelve-month data-center revenue estimates and hyperscaler capex guidance remain intact. Use a 3-6 month NVDA call spread rather than outright leveraged ETF exposure; target roughly 2:1 upside/downside and exit if forward EPS estimates fall more than 5%.
• Express the likely dispersion trade over 1-3 months: long NVDA / short PLTR in dollar-neutral sizing. NVDA has more near-term earnings support, while PLTR is more vulnerable to duration-driven multiple compression; close if PLTR's commercial revenue growth reaccelerates enough to lift consensus revenue estimates by more than 10% or if NVDA guides margins materially lower.
• Avoid long NVDL and similar daily-reset leveraged products as strategic holdings. For accounts able to borrow reliably, a tactical short NVDL against a smaller long NVDA hedge can monetize volatility drag during a sideways or choppy market; treat as a days-to-weeks trade and cap risk because a sharp NVDA rally can overwhelm reset decay.
• Set a forced-selling watchlist rather than initiating broad shorts: monitor FINRA margin-debt releases, NVDA/TSLA implied volatility, and high-yield credit spreads. If volatility rises while spreads stay contained and earnings estimates hold, use the dislocation to add quality AI exposure; if spreads widen materially alongside estimate cuts, reduce beta through SOXX or QQQ hedges.