AI Bubble Fear: Burry Shorts Nvidia, Palantir, Tesla as Margin Debt Up 50%

For AI builders and infrastructure teams, Burry's short positions are a signal about capital flow fragility more than model capability. Nvidia remains the compute pillar for LLM training, Palantir is scaling enterprise AI deployment, and Tesla is pushing autonomous and embodied AI, but a forced deleveraging could stall the very funding that accelerates these roadmaps.

Michael Burry, founder of Scion Asset Management, has maintained short positions in Nvidia, Palantir, and Tesla while warning that artificial-intelligence equities could suffer a "1987-type fall," according to a September 12, 2026, commentary by The Motley Fool syndicated on AOL. The article reports Burry articulated his concerns in a Substack newsletter, focusing not solely on valuations but on the leverage embedded in today's market structure. The precise timing of the newsletter isn't given, but the commentary arrived as margin debt had already surged 50% year over year as of June 2026, giving a concrete macro data point to the argument that the current AI-led rally is increasingly debt-financed.

“

Nvidia remains the compute pillar for LLM training, Palantir is scaling enterprise AI deployment, and Tesla is pushing autonomous and embodied AI, but a forced deleveraging could stall the very funding that accelerates these roadmaps.

Burry's concern revolves around the reflexive mechanics of leveraged funds. The analysis cites the GraniteShares 2x Long NVDA Daily ETF (NVDL) as an example: because the fund must maintain two times daily exposure, it is forced to buy more Nvidia shares as the stock rallies, adding fuel to the rally. That dynamic works in reverse in a selloff, creating forced selling that can cascade. The article also points to margin debt growing 50% year over year in June as evidence that individual and institutional investors are borrowing against positions to amplify gains. The article cites the liquidation of Leopold Aschenbrenner's Situation Awareness hedge fund as a demonstration of downside risk. The fund had deployed highly leveraged bets on hot AI stocks; when prices declined, forced deleveraging led to liquidation. That real-world example, the author notes, makes Burry's cautionary point about leverage "valid." For leveraged investors, even a 20% drawdown can be devastating, while long-term unleveraged investors can usually recover.

However, the article splits Burry's argument: his leverage concern is credited, but his AI bubble thesis is challenged. The author argues that Nvidia, Palantir, and Tesla are backed by strong fundamentals, multiyear tailwinds, and rising profit margins, which historically do not align with a bearish outcome over long horizons. It notes Burry has made several bad calls, especially for long-term investors, and that his bearish positions often use puts, which are short-term instruments. That matters: a put expiring in a matter of months can be burned by rallies even if a correction eventually arrives. Thus, readers are reminded that Burry's trade is a timing bet, not necessarily a multiyear valuation call.

From an AI-industry perspective, the stakes are high. Nvidia's chips remain the compute backbone for large-language-model training; Palantir is expanding in government and commercial AI deployment; Tesla's AI ambitions span from autonomous driving to robotics. A 1987-style event — in which the S&P 500 fell about 22.6% in one day on October 19, 1987 — would not necessarily invalidate those product roadmaps, but it could freeze the capital that fuels their expansion. AI startups depending on continuous funding rounds or contractual compute purchases would feel the pressure, and even chip buyers may delay purchases if equity holdings collateralize their capital. The article's mention of leveraged ETFs like NVDL also highlights how retail and hedge-fund flows are increasingly entwined with the most prominent AI names, which could amplify both rallies and corrections well beyond changes in underlying earnings.