Burry reads Big Tech’s AI spending as a bid to become too big to fail

Michael Burry says Big Tech chief executives are betting on a “too big to fail” artificial intelligence oligopoly^5,6. He made the claim in a post on X, where he described the spending as a bid to be “too big [to] fail”³. The argument, as summarised by the outlets carrying it, is that the companies are spending for oligopoly status rather than for a near-term return^3,5. Outlets identify him as the “Big Short” investor².

Burry has recently argued that the tech giants, with the exception of Apple, “are spending as if their lives depended on it”¹. He singled out Microsoft, saying it uses spending metrics “as a lever” to get its way, and questioned its AI and data-centre spending while warning of “too big to fail” oligopoly motives^1,7.

Who else he puts in the frame

The claim extends past the incumbents. Burry says OpenAI and Anthropic want to join Big Tech’s oligopoly, in his account of how the AI leaders could become a group too important to fail^2,4.

He also put a label on the market’s own state, saying it is in a “denial” phase like the ones seen in 2000 and 2008, and flagging risks to data-centre financing⁴. One account of his remarks paired them with a sceptic’s view that the companies have “no hypergrowth ideas left”^5,6. The oligopoly claim and the denial claim were carried by several outlets within a day of each other^3,4,5,6.

What it means for investors

The reading on offer treats the spending as a bid for implicit support rather than as a project with a measurable return, which is a different question from whether the outlay is affordable^3,7. It also puts the motive, not the amount, at the centre of the case: Microsoft’s use of spending metrics “as a lever” is the specific conduct Burry names¹.

The sources record a thesis, not evidence. They give no figure for the spending, no data-centre contract, no position Burry holds, and no response from Microsoft, OpenAI or Anthropic^1,2,7. Analysts Joachim Klement and Francisca Reis appear in the same coverage, though the excerpts do not carry their conclusion⁴. What is on the record is the comparison to two previous cycles and a named risk, the financing of data centres, that the coverage attaches to it⁴. Nor do the sources record a date for the posts, beyond their circulation in the coverage over the past two days^3,5.