Inside Anthropic’s $1 Trillion Spending Plan On Data Centers, AI Chips

Anthropic’s IPO prospectus filed on Monday unveiled the enormous scale of its spending on AI chips and data centers. The fast growing firm, which has already spent over $19 billion on AI infrastructure since its founding in 2021, has already signed contracts totaling more than half a trillion dollars over the next 20 years.

Those include deals with major hyperscalers like Amazon, Google and Microsoft as well as chipmakers AMD, Broadcom and Nvidia plus compute deals with Elon Musk’s SpaceX and a host of neoclouds and data center operators including CoreWeave, Google-backed Fluidstack, Nvidia-backed Lambda and Nscale. Much of that was already public, but the size of the contracts wasn’t until now.

According to the prospectus, Anthropic was on the hook for approximately $587 billion in chip and data center commitments through 2045 as of June. Then, over the past four months, Claude’s maker signed another $266 billion in new and amended agreements. There’s also $20 billion in expected computing capacity from AMD and a further $125 to $150 billion for more Google-designed chips. That means the company’s obligations could total around $1 trillion through 2045—about 3% of the total $31.6 trillion that consulting giant PwC estimates will be spent on the global AI buildout by 2050. Anthropic’s biggest rival, OpenAI, is reportedly budgeting $750 billion for computing power through 2030.

It’s a massive step up compared to what Anthropic has spent in its short history. From January 2023 to June 2026, the company paid $19.2 billion on AI compute and infrastructure, including data centers, chips, equipment and software to manage those operations. Over that same time period, the firm generated $21.3 billion in revenues. The majority of that revenue, $16.3 billion, was recorded in just the first half of 2026, a 1,300% jump year-over-year while compute and infrastructure expenses increased by just 283%.