Anthropic's Cloud Partners Are Starting to Look Like Rivals

Anthropic's biggest cloud suppliers are starting to act like competitors, and that's creating a capacity problem the company can't easily solve with money alone.

A leaked S-1 filing, reported by The Wall Street Journal, shows the Claude-maker struggling to lock in enough computing power from AWS, Google, and Microsoft. The issue isn't budget. Anthropic says it could spend more than $750 billion on cloud contracts, some of them still being negotiated. The issue is that the hyperscalers it depends on for that computing power increasingly view Anthropic as a rival AI player rather than just another customer buying server time.

The filing adds detail to numbers Reuters and the Financial Times first reported in late September, and it paints a picture of a company spending at an extraordinary rate while trying to secure infrastructure that its own suppliers have growing reasons to ration.

The workaround, and its risks

Anthropic's response has been to diversify away from the big three. The company is leasing its own data centers and striking deals with third-party operators outside the traditional hyperscaler circle, according to the filing. But that comes with a tradeoff Anthropic itself flags: many of these smaller operators have what the filing calls a "limited operating history," meaning they haven't been running at scale long enough to prove they're as dependable as AWS, Google, or Microsoft.

That's a real bet. If Anthropic's AI ambitions depend on massive, reliable compute, outsourcing a growing share of that to newer, less-tested operators introduces risk the company wouldn't face if its main partners weren't simultaneously building competing AI products.

The money problem underneath it all

The capacity squeeze is happening against a backdrop of staggering losses. Anthropic reported an $80.7 billion net loss in the first half of the year, against $16.3 billion in revenue over the same period, per the WSJ report. Just two customers accounted for roughly a fifth of that revenue, a concentration that underscores how much the company's financial picture still rests on a handful of relationships.

There's a growth story in there too. Anthropic told prospective investors last week that third-quarter revenue hit $16.8 billion, up 45% from the prior quarter. That's a sharp trajectory, even if it does little to offset losses of that magnitude.

Liability on top of spend

The filing also expands on catastrophic-risk disclosures Anthropic had previously made public. It warns that if the company's AI systems cause harm, Anthropic could be on the hook for damages that exceed what its insurance would actually cover. For a company building increasingly powerful AI models and racing to scale them, that's a liability exposure that compounds the financial risk already visible in its burn rate.

Why this matters

Anthropic is reportedly aiming for a November IPO, and S-1 filings exist to give investors the real picture before money changes hands. What this one shows is a company caught in a structural bind: it needs enormous, reliable compute to compete in AI, but the companies best positioned to provide that compute are themselves building competing AI products. That tension isn't unique to Anthropic. OpenAI, Google, and Microsoft are all tangled in similar cross-cutting relationships where cloud providers, model developers, and AI competitors increasingly overlap. But Anthropic's filing puts dollar figures on just how expensive and precarious that tension has become for one of the industry's most prominent players.

Anthropic did not respond to TechCrunch's request for comment on the filing.