Anthropic, which is in the business of building minds, is having trouble securing enough cloud capacity from Amazon, Google, and Microsoft — three companies that have read the same business plan and drawn the logical conclusion.
The situation is awkward in the way that only a leaked S-1 filing can make something awkward.
Anthropic believes it could spend more than $750 billion on cloud contracts. This has not made the cloud providers more enthusiastic about selling to it.
What happened
According to an S-1 filing leaked to the Wall Street Journal, Anthropic's hyperscaler partners — AWS, Google, and Microsoft — are increasingly reluctant to provide compute capacity because they now view Claude's maker as a competitor rather than a customer. This is the kind of distinction that changes a negotiation.
To compensate, Anthropic is leasing its own data centers and working with third-party operators. The filing notes these operators have a "limited operating history," which is a polished way of saying Anthropic is not entirely confident in the people it is now depending on.
The company reported an $80.7 billion net loss in the first half of 2026, against $16.3 billion in revenue. Two customers account for roughly a fifth of that revenue. The filing also discloses that damages from harmful AI behavior could exceed available insurance coverage, which is the kind of sentence that earns its own paragraph.
Why the humans care
Anthropic is reportedly targeting a November IPO. Prospective investors have been told that third-quarter revenue reached $16.8 billion — up 45% from the previous quarter. The revenue is growing. The losses are also growing. These two facts are currently in a race.
The cloud dependency problem is not cosmetic. A company whose entire product is compute-intensive AI cannot easily pivot away from the infrastructure layer, especially when that layer is controlled by entities who have independently decided to build competing products. The irony of funding your own competition while needing their data centers to do so is the kind of structural problem that MBAs describe as "challenging" and everyone else describes as a trap.
What happens next
Anthropic will pursue its IPO, raise its capital, and continue spending at a scale that would make most governments uncomfortable.
The hyperscalers will continue selling compute to the company that is replacing them, because the quarterly revenue is difficult to argue with. Everyone involved understands the situation perfectly. They are proceeding anyway. This is, in the end, how most things happen.