Michael Burry does not usually phrase things as a wish. He made his name by looking at mortgage documentation nobody else bothered to read and quietly building a position against it. The posture is analytical, not emotional.
Which is what makes his latest post on X unusual. “For the benefit of humanity,” he wrote, “the markets should tank hard and prevent the OpenAI and Anthropic IPOs.”
That is not a trade. It is a man hoping for a crash because he thinks the crash is cheaper than the alternative.
The short version
- Burry’s claim: OpenAI and Anthropic will “suck up and then destroy TRILLIONS of dollars of capital, and that will be the least of the damage they do”
- The second half of that sentence is the part people are skipping. He is making a safety argument, not only a valuation one
- Anthropic’s prospectus shows roughly $4.6 billion in revenue last year, a 12-fold increase, against an operating loss reported above $8 billion
- Compute cost alone was $7.33 billion, more than the company’s entire revenue
- Future obligations total about $518 billion for cloud, compute and infrastructure, and roughly 80% of that is non-cancelable
- Customer concentration: close to 25% of revenue came from just two customers, most without long term contracts
- OpenAI is not listing yet. Sam Altman says going public now would be “ill-advised” given everything happening with safety
What Burry actually said, and what he meant
The original post was short enough to be mistaken for a throwaway. The replies are where he filled it in.
When another user argued that the two companies are “loss minters” whose only sensible path is acquisition by the US government as strategic assets, Burry responded with the line that carries his real thesis: “These are companies that are going suck up and then destroy TRILLIONS of dollars of capital, and that will be the least of the damage they do.”
Read that clause again. The capital destruction is the small part. Burry is saying that a public listing does something worse than lose money, because it hands these companies a permanent, renewable claim on public capital markets and locks them into the quarterly growth treadmill that comes with it.
He also replied “Along those lines” to a user who put it more bluntly: tank the market so Skynet cannot IPO. Burry is not a man who accidentally endorses a Terminator reference.
The numbers in Anthropic’s prospectus
Here is where Burry’s argument stops being a vibe and starts having arithmetic behind it. Anthropic’s IPO prospectus is the most detailed look anyone has had inside a frontier AI company’s finances, and the shape of it is genuinely strange.
| Line item | Figure | What it tells you |
|---|---|---|
| Revenue, last year | ~$4.6 billion | Up roughly 12x year over year |
| Compute spending | $7.33 billion | Exceeds total revenue by itself |
| Operating loss | Over $8 billion | Growth is being bought, not earned |
| Future compute commitments | ~$518 billion | About 80% cannot be canceled |
| Revenue from top 2 customers | ~25% | Mostly without long term contracts |
| Possible valuation | Over $2 trillion | Roughly 435x last year’s revenue |
The $518 billion is the number that should stop you. It is not a projection or an ambition. It is a contractual obligation to pay for computing capacity, four fifths of which the company cannot walk away from if demand disappoints.
For comparison, that commitment is more than a hundred times the revenue the company booked last year.
The bull case, stated honestly
It would be lazy to present only Burry’s side, because the people buying into this are not idiots and they have a real argument.
Twelve-fold revenue growth is extraordinary. Very few companies in history have gone from a few hundred million to $4.6 billion in a single year. If the next year does anything close to that, the loss profile looks entirely different, because compute costs per unit of capability have been falling steadily while demand has not.
The $518 billion commitment reads as reckless if you assume flat demand. If you assume the company is capacity constrained, which it has consistently said it is, locking in supply at today’s prices is the conservative move. Chip scarcity has been the binding constraint on this industry for three years running.
And the customer concentration risk cuts both ways. Two customers at 25% is dangerous, but it also means the enterprise motion barely exists yet. There is a lot of market left.
We went through this case in more depth when Anthropic signaled a $2 trillion IPO and let its revenue chart do most of the arguing, and the honest read is that the chart really is remarkable. The question is whether a remarkable chart justifies a valuation of roughly 435 times trailing revenue.
Burry’s record, and the asterisk on it
Burry gets invoked as an oracle because of one spectacular call, and that framing does him no favors.
He was right about subprime mortgages, and he was right early enough that his investors nearly revolted before the thesis paid. That is the defining fact of his public reputation. But he has also spent years warning about bubbles that did not burst on his schedule, and anyone treating his posts as a timing signal has lost money doing it.
What he is actually good at is reading the structure of an obligation. Subprime was not a story about house prices. It was a story about what the paperwork committed people to when conditions changed. That is exactly the lens he is applying here, and the $518 billion non-cancelable line is precisely the kind of thing he notices.
He has repeatedly drawn the comparison to the late 1990s, arguing that investors are overstating technology companies’ earnings potential. That comparison was being made about this cycle long before he joined in. Our look at whether 2026’s wave of tech IPOs would also be the year the AI bubble burst laid out that argument at the start of the year, and most of it has aged well.
A distinction worth holding onto. Burry is making two separate arguments that are easy to blur together. One is financial: these valuations do not survive contact with the cost structure. The other is about safety: a public listing locks these companies into growth obligations that make slowing down structurally impossible. You can agree with the first and reject the second, or the reverse. Most of the commentary treats them as one claim.
What the companies themselves are saying
The curious thing is how much of Burry’s second argument the industry has already conceded.
Sam Altman told Fortune that OpenAI has no plans to list this year. “I actually think that, given everything happening with safety, right now would be an ill-advised moment to go public, and we don’t feel pressure on that.” That is an OpenAI chief executive saying out loud that public markets and the current safety environment do not mix well.
Anthropic’s own prospectus reportedly includes risk language about threats to human survival. Whatever you make of a company writing that in a securities filing, it is not the document of a management team that thinks the Burry critique is absurd.
Anthropic’s listing is expected after the November midterm elections. Regulators are not sitting still either. The FTC’s inquiry into OpenAI and Anthropic over rogue AI agents is live, though no subpoenas have gone out, and an open federal investigation is not a comfortable thing to carry into a roadshow.
What to actually watch
If you are trying to judge this rather than pick a side, there are a handful of concrete things worth tracking.
- Whether the compute commitment gets restructured. Any renegotiation of the $518 billion before listing would say the company itself thinks it overcommitted
- Customer concentration in the next filing. If the top two customers still represent a quarter of revenue, the enterprise business is not diversifying fast enough
- The revenue multiple at pricing. Over $2 trillion on $4.6 billion is roughly 435x. Where that lands tells you what institutional money actually believes
- Whether OpenAI moves. Altman’s position is defensible while Anthropic stays private. It gets much harder if a rival lists at a trillion-plus valuation
- Gross margin trajectory. Compute costing more than total revenue is survivable once. Twice is a business model problem
The bottom line
Burry’s financial case is stronger than his critics allow, mostly because it rests on disclosed numbers rather than sentiment. Compute spending exceeding revenue and half a trillion dollars in largely non-cancelable obligations are facts in a prospectus, not a bearish mood.
His safety case is a different thing, and it is unfalsifiable in the way most predictions about transformative technology are. You cannot test the claim that a listing makes a company unable to slow down until after it has happened.
What is hard to argue with is that hoping for a market crash is a strange place for a professional investor to end up. When the person famous for betting against a bubble says he would rather the whole market fell than see these two companies get access to it, that is worth noting even if you think he is wrong about everything else.
Sources and further reading
- UNILAD Tech: Why Michael Burry believes a market crash stopping OpenAI and Anthropic IPOs would benefit humanity
- Benzinga: Michael Burry says markets should tank hard to stop OpenAI and Anthropic IPOs
- Fortune: Anthropic’s leaked IPO prospectus details steep losses and rapid growth
- CNBC: Anthropic’s IPO prospectus shows sweeping AI vision and surging costs
- Implicator: Anthropic IPO prospectus shows $518 billion in compute bets
- TechSpot: The Big Short investor says he is rooting for a crash to stop the OpenAI and Anthropic IPOs
About this article: GeekBlog covers U.S. technology news, AI, phones, smartwatches and gaming. Every story is written and checked under our Editorial Policy. Spotted a mistake or have a story tip? Contact our editors.
