What Burry actually said
In a post on X, Burry wrote that markets should "tank hard" in order to stop the planned public offerings of Sam Altman's OpenAI and of Anthropic, and called that outcome one that would be "for the benefit of humanity."
He didn't stop there. He went on to claim the two companies would "suck up" and then destroy "TRILLIONS of dollars of capital," and that the money investors stood to lose would be "the least of the damage they do."
Coming from most people that reads as noise. Coming from Burry, who has a habit of being early and right about things nobody wants to hear, it gets attention.
The money behind the warning
The timing isn't random, because both companies really are inching toward the public markets with enormous bills attached. A confidential Anthropic IPO prospectus is reported to show at least $518 billion in future cloud, computing and infrastructure commitments, with the company possibly chasing a valuation north of $2 trillion when it lists.
OpenAI quietly filed its own draft registration back in June, though Altman has said now isn't the right moment and the company is reportedly eyeing a 2027 debut instead. Burry's core worry sits underneath those numbers.
He sees a gap between the staggering amount of money being poured in and how unproven the actual economics are, and he keeps asking who eats the losses if AI revenue never grows fast enough to cover the infrastructure it runs on.
Not his first AI alarm
This fits a pattern he's been building for about a year, as he's turned steadily more bearish on the whole sector. His bets have reportedly lined up against Nvidia, Palantir, Micron, Oracle and the Nasdaq 100, all resting on a single idea, that borrowing huge sums to build out chips and data centres can't go on forever and would crack fast if interest rates climb.
The X posts came the same day as a Substack newsletter in which he said he's more sure than ever that his thesis plays out within a year, pulling his own call forward from a previous base case that had pegged 2028 for the bubble to pop.
Even the labs flag the danger
Here's the part that cuts both ways. OpenAI and Anthropic have each admitted, at least on paper, that their technology carries real dangers, and Anthropic's draft prospectus reportedly goes as far as saying its systems could pose catastrophic or even existential risk to humanity.
That sounds like it backs Burry up, but it's worth reading the fine print, because that kind of language is standard in the risk-factor section of any IPO filing, written to warn investors of what could go wrong rather than to predict that it will.
So the warning and the disclosure aren't quite the same thing. A famous bear can move a lot of money with a single post, but wanting a crash is not the same as one arriving.