Anthropic plans to spend at least $518 billion on AI infrastructure over roughly a decade, according to a confidential IPO prospectus reviewed by Reuters. The scale of that figure puts a number on something the industry has long suspected: running a frontier AI lab now costs as much as building a small country’s infrastructure.
The commitments span cloud computing, specialized AI hardware and long-term agreements with Amazon, Google, Microsoft and Broadcom. Notably, Reuters reported that roughly 80 percent of these commitments are non-cancellable or require payment regardless of actual usage, meaning Anthropic is locked into most of this spending no matter how its business evolves.
Breaking down the numbers, Anthropic’s commitments include $111.1 billion with Google, $110 billion with Amazon and $31.4 billion with Microsoft, alongside $161.2 billion in largely non-cancellable equipment leases tied to Broadcom. For context, the company spent more than $7 billion on compute and infrastructure in 2025 alone, while its operating loss exceeded $8 billion.
Amazon remains Anthropic’s primary cloud and training partner. In April, the companies expanded their agreement so Anthropic would commit more than $100 billion over ten years to AWS, securing up to 5 gigawatts of additional computing capacity. At the time, Anthropic said it was already using more than one million Trainium2 chips to train and serve Claude.
Google plays a similarly central role. Anthropic said in October 2025 that it planned to expand its use of Google Cloud to as many as one million TPUs, a deal worth tens of billions of dollars. That relationship grew further in April through a joint agreement with Google and Broadcom covering next-generation TPU capacity, which Anthropic later described as representing 5 gigawatts.
Microsoft and Nvidia round out the picture. Anthropic committed to $30 billion of Azure compute capacity in November 2025, while Nvidia agreed to invest up to $10 billion in Anthropic and Microsoft up to $5 billion as part of the same deal.
Despite the infrastructure buildout, Anthropic’s revenue has grown quickly alongside it, reaching an annualized run rate above $30 billion in April and crossing $47 billion by May. Even so, the company flagged its dependence on major cloud partners as a business risk in its own filing, a reminder that scaling AI still means betting heavily on someone else’s data centers.
