Big Short's Michael Burry Switches His AI Trade as He Positions for a Bubble That 'May Burst' Sooner

Michael Burry has shifted his high‑profile bet against the artificial intelligence boom into long‑dated options, warning the 'bubble in AI may burst sooner than later' after new research prompted him to bring forward his timeline.

In a 28 September post on his Cassandra Unchained Substack, the investor, who became known for betting against the US housing market before the 2008 financial crisis, said he had replaced a series of outright short positions with put options.

He said the move would give him more leverage over a shorter period because 'the bubble in AI may burst sooner than later'.

Burry said the change was designed to give him more leverage over a shorter period. He also said tax‑loss harvesting played a secondary role in some transactions, while most of the repositioning followed research he conducted over the weekend.

Burry Switches Shorts for Puts Running Into 2027

Burry replaced or rolled several outright short positions into dated puts, with contracts extending into 2027. For Micron, he bought June 2027 puts around the $500 strike. He also bought June 2027 puts on Nebius in the double‑digit strikes and September 2027 puts on the iShares Semiconductor ETF, or SOXX, in the low $400s.

His Palantir position was enlarged and rolled into September 2027 puts in the low $100s. He replaced his Nvidia short with September 2027 puts in the mid‑$100s.

Burry also replaced his Caterpillar short with December 2027 puts around $400 and his Oracle short with December 2027 puts in the mid‑double digits. His Nasdaq exposure was rolled into September 2027 Nasdaq‑100 puts in the $24,000s. He opened new long‑dated puts on MetLife in the $70s.

CoreWeave was an exception. Burry covered the short but did not buy puts because he said the available prices were unattractive. Puts expire. Shorts do not.

Research Moves Burry's AI Timeline Forward

His previous base case was 2028. In his 28 September post, Burry said new research had caused him to 'move timelines up'. He separately highlighted an Ares Management analysis of financing behind the digital‑infrastructure build‑out, saying its work lined up with his own research.

Burry said Ares documented 26 disclosed financings worth about $573 billion over the previous 12 months, alongside more than 100 digital‑infrastructure financings in total. He said roughly half of the broader group was private.

The transactions included exposure to major technology companies including Meta, Oracle, Microsoft, Amazon, Google and Nvidia, with exposure linked to OpenAI and Anthropic.

Ares examined how financing structures supporting AI infrastructure depend on continued capital spending. It warned that a period of disappointing AI revenue could cause companies to reconsider spending, potentially putting guarantees and other contractual obligations under pressure if guarantors themselves faced financial strain.

Ares was analysing concentration in the financing and the assumptions supporting the build‑out, rather than forecasting an AI collapse.

The Strikes Are Not Price Targets

The scale of Burry's strikes is visible against the 28 September closes. Micron finished at about $1,054, compared with the roughly $500 strike on his June 2027 puts. Palantir closed near $187, while the puts he disclosed were in the low $100s. SOXX ended around $561 against puts in the low $400s.

Those strikes are contract terms, not price targets. The options must gain value before their respective expiry dates for the trades to deliver the intended downside exposure. Burry also said volatility remained subdued, making some of the puts cheaper than he believed they should be.

Burry added shares of Sprouts Farmers Market in the $63s and bought more QXO in the low $12s. QXO is a building‑products distributor led by Brad Jacobs.