Leaked S-1 shows Anthropic short of cloud capacity as AWS, Google and Microsoft turn rivals

Anthropic is struggling to secure enough cloud computing capacity from AWS, Google and Microsoft, three suppliers that increasingly see the maker of Claude as a rival rather than a customer, according to an S-1 filing leaked to The Wall Street Journal and reported by TechCrunch.

The leaked paperwork adds detail to figures that Reuters and the Financial Times first surfaced in late September, and it arrives as the company is reportedly aiming for a November IPO.

The scale of the commitment

In the filing, Anthropic estimates it could spend more than $750 billion on cloud contracts, a total that includes deals still being negotiated, TechCrunch reports. Even commitments on that scale, however, have not guaranteed the compute the company needs to keep up with demand for its models.

Working around the hyperscalers

To close the gap, Anthropic has begun leasing its own data centers and contracting with third-party data center operators. The filing concedes a drawback of that strategy as well: the company is concerned these operators, many of them relatively new to the business, may prove less dependable than the established cloud giants they are meant to supplement.

Losses, revenue and customer concentration

The filing quantifies the financial backdrop to all that spending. Anthropic posted an $80.7 billion net loss in the first half of the year against $16.3 billion in revenue, according to the WSJ's account, and roughly a fifth of those sales came from just two customers.

More recent figures point to steep growth. TechCrunch reports that Anthropic told prospective investors last week that third-quarter revenue alone reached $16.8 billion, up 45% from the previous quarter.

Liability beyond insurance

The document also builds on previously reported disclosures about catastrophic risks. It warns that Anthropic could be held liable for damages caused by harmful AI behavior at a level that would exceed the insurance coverage the company has in place.

Anthropic did not immediately respond to TechCrunch's request for comment.

Why it matters

The filing documents a turning point in the AI supply chain: the major cloud providers are no longer neutral landlords. If AWS, Google and Microsoft treat one of their largest compute buyers as a competitor, access to the infrastructure AI products are built on becomes a competitive factor in itself. For a compute-bound business like Anthropic, dependence on firms it competes with is not a procurement nuisance but a structural constraint on growth.

The leak also sharpens the picture for anyone weighing the reported November listing. A prospective investor has to balance fast-rising revenue against an $80.7 billion half-year loss, heavy reliance on two customers for a fifth of sales, and the possibility that capacity may not be available from the hyperscalers at any price the company can afford. Anthropic's response, leasing its own facilities and turning to newer operators, suggests the largest AI labs will increasingly own their infrastructure, taking on capital costs and reliability risks that the hyperscalers were originally supposed to absorb.