Reuters reported Tuesday that Anthropic expects to spend at least $518 billion over a decade building AI infrastructure with six partners, citing a confidential IPO prospectus it reviewed. The plan ranks among the largest AI buildout commitments on record — and about 80% of the sum is non-cancelable or payable regardless of usage.

That last detail is the story. These aren't purchase orders a CFO can quietly trim in a downturn; they're decade-long take-or-pay obligations. Anthropic tells investors the commitments are necessary because access to computing power is becoming the key constraint on AI development, with future demand for advanced systems "limited principally by the availability of compute." The company filed for an IPO confidentially with the Securities and Exchange Commission in June, but its paperwork has not been publicly disclosed. Anthropic did not respond to Reuters' request for comment.

The take-or-pay structure#

The prospectus breaks the spend into long-term infrastructure service obligations running seven to ten years, "regardless of usage": at least $111.1 billion with Alphabet's Google between April 2026 and July 2033, $110 billion with Amazon between May 2026 and April 2036, and $31.4 billion with Microsoft between November 2026 and May 2033. The terms are blunt — "If our actual spend falls short, we must pay Google the difference," the company says, adding that similar terms apply to Amazon. The Microsoft commitment is described as non-cancelable except in the event of Microsoft's uncured material breach.

Close-up of AI server hardware with glowing status lights
AI server hardware — the asset Anthropic is pre-buying a decade of access to. Illustration: AI Frontier Post.

On top of the cloud deals, Anthropic is carrying about $161.2 billion of Broadcom-related equipment lease obligations, largely non-cancelable by either party except in case of default. Six partners total share the $518 billion envelope. In other words, the lab has effectively signed a mortgage on the next ten years of compute — before a single share of the IPO has priced.

The financials behind the filing#

The prospectus shows a company burning cash at frontier-model speed. Anthropic posted a net loss of $42 billion in 2025, though roughly $34 billion of that was an accounting charge reflecting the increased estimated value of financing that could eventually convert into shares — not cash spent running the business. On an operating basis, the loss widened to $8.06 billion from $2.98 billion in 2024. Revenue grew twelvefold to nearly $4.6 billion, while compute and infrastructure spending tripled to $7.33 billion.

Aerial night view of a hyperscale data center campus
A hyperscale data center campus at night — Anthropic is locking in a decade of this kind of capacity. Illustration: AI Frontier Post.

Two risk factors stand out. Nearly a quarter of 2025 revenue came from just two customers, and the prospectus warns that many of its largest clients are not locked into long-term contracts. The company held $20.28 billion in cash and short-term investments at the end of 2025. Reuters puts a potential public valuation above $2 trillion, more than double the $965 billion valuation Anthropic announced in its May 2026 funding round — with the debut likely pushed past the November midterm elections.

Why it matters now#

The Anthropic filing is the demand-side mirror of the week's infrastructure arms race. On the supply side, suppliers are inventing new financial structures to keep the buildout funded; on the demand side, labs are pre-buying a decade of compute in contracts that don't cancel. That works beautifully if model revenue keeps compounding. If it doesn't, the prospectus makes the downside explicit: the bills arrive regardless of usage.

Sources