Anthropic’s IPO Papers Show $518 Billion in Future Compute Costs

About 80% of Anthropic’s planned $518 billion infrastructure spending cannot be canceled or must be paid regardless of usage, Reuters reports from its confidential IPO prospectus. The commitments span a decade and six partners. That makes much of the bill an obligation, rather than spending Anthropic could simply trim if demand disappoints.

The filing lists minimum payments of $111.1 billion to Google, $110 billion to Amazon and $31.4 billion to Microsoft over seven to ten years, regardless of usage. Another $161.2 billion covers Broadcom-related equipment leases that are largely non-cancelable. Under the Google agreement, Anthropic must pay the difference if actual spending falls short; Amazon has similar terms.

There is more flexibility elsewhere: agreements with xAI could involve up to $84.5 billion through 2029, but are largely cancelable with 90 days’ notice.

Anthropic argues computing capacity will constrain future AI demand. Securing it now protects access, but commits it to paying before knowing how much it will need. Anthropic did not immediately respond to Reuters’s request for comment; the prospectus remains confidential.

Nearly a quarter of Anthropic’s 2025 revenue came from just two customers, according to the IPO prospectus reviewed by Reuters. The company also warned that many of its largest clients have no long-term contracts and could reduce or stop spending. That adds a customer-retention risk to the enormous computing commitments disclosed in the same document. The concentration figure describes last year, not today’s customer mix.

A separate Reuters report details how control would remain with the founders after the offering. A majority of the seven cofounders would direct a single Class F share carrying 50.1% of voting power on key shareholder matters through a new “Founder LLC.” The Long-Term Benefit Trust would still elect four of the seven board directors.

Public investors would therefore be buying into the business without gaining control of it. The prospectus warns that the governance structure could produce decisions contrary to shareholders’ financial interests. Anthropic declined to comment to Reuters.