NEW YORK CITY, New York: Anthropic expects to spend at least US$518 billion on artificial intelligence infrastructure over the next decade, with most of its planned commitments requiring payment even if the company does not use the computing capacity, according to a confidential IPO prospectus seen by Reuters.

The AI company said about 80 percent of the commitments across six partners are non-cancelable or payable regardless of usage. It told prospective investors that access to computing power is becoming a key constraint on AI development, with future demand for advanced systems likely to exceed supply and be "limited principally by the availability of compute."

Anthropic plans to spend at least $111.1 billion with Alphabet's Google, $110 billion with Amazon and $31.4 billion with Microsoft under long-term infrastructure agreements spanning seven to 10 years.

The company said its Google and Amazon agreements require minimum payments even if actual spending falls short.

"If our actual spend falls short, we must pay Google the difference," Anthropic said, adding that similar terms apply to its Amazon agreement.

Its Microsoft commitment runs from November 2026 to May 2033 and "is non-cancelable except in the event of Microsoft's uncured material breach," according to the prospectus.

Anthropic also disclosed about $161.2 billion in Broadcom-related equipment lease obligations that are largely non-cancelable. The company said those arrangements cannot be canceled by either party except in the case of a default.

Other agreements have different terms. Anthropic disclosed arrangements with Elon Musk's xAI that could result in up to $84.5 billion in spending on Nvidia-based computing capacity through 2029. Those commitments are largely cancelable with 90 days' notice.

The filing also showed that Advanced Micro Devices has committed to buying up to $5 billion of Anthropic stock and supplying AI computing capacity expected to exceed $20 billion.

Anthropic said it is moving beyond a cloud-only approach by building dedicated data centers and directly leasing chips. It identified a risk in its reliance on Amazon, Google and Microsoft, which supply computing infrastructure and distribution while also developing competing AI models.

The three companies serve as investors, customers, cloud providers, distributors and competitors, creating incentives that "may not be fully aligned" with Anthropic's interests, the prospectus said.

"If the compute we have access to from third parties is curtailed, repriced, or terminated ... our business, financial condition, and results of operations could be adversely affected," Anthropic said.

The company confidentially filed for an IPO with the U.S. Securities and Exchange Commission in June. Its prospectus has not been publicly disclosed, and Anthropic did not immediately respond to a request for comment about the filing.