Anthropic plans over $518 billion for AI infrastructure — 80% of commitments irreversible

Artificial-intelligence company Anthropic has told investors it expects to spend at least $518 billion over a decade building AI infrastructure with six partners — one of the largest AI build-out commitments ever recorded, according to a confidential IPO prospectus reviewed by Reuters.

Under the document, around 80% of that sum is irreversible or requires payment regardless of usage. Anthropic tells investors the commitments are necessary because access to computing power is becoming the main bottleneck in AI development, while future demand for advanced AI systems is likely to outstrip available supply.

The company plans to spend at least $111.1 billion with Google, $110 billion with Amazon and $31.4 billion with Microsoft under long-term infrastructure-service obligations over the next 7 to 10 years, “regardless of usage,” as Reuters reports. Anthropic also carries about $161.2 billion in equipment-lease obligations with Broadcom that are mostly irreversible.

The document also reveals an agreement with Elon Musk’s xAI that could result in up to $84.5 billion in spending on Nvidia-chip-based computing capacity by 2029 — mostly cancellable with 90 days’ notice — as well as a deepening relationship with AMD, which has committed to buying up to $5 billion of Anthropic stock and supplying computing capacity expected to exceed $20 billion.

The total is comparable in scale to OpenAI’s Stargate project, the $500 billion AI-infrastructure plan. Anthropic says it is increasingly building its own infrastructure, moving from a cloud-only model toward dedicated data centers and directly leased chips — and warns of dependence on Amazon, Google and Microsoft, which are simultaneously investors, cloud providers and competitors with their own models. The company filed confidentially for an IPO with the SEC in June, while the agency reports a targeted valuation above $2 trillion.