Anthropic to spend $100B on AWS compute over 10 years

What happened

Anthropic announced it will spend over $100 billion（約16兆円） on Amazon Web Services over the next 10 years, including up to 5 gigawatts of compute capacity. The company also uses over 1 million Amazon Trainium 2 chips and up to 1 million of Alphabet's Tensor Processing Units (TPUs), and relies on Nvidia's GPUs as well. Anthropic's annualized revenue has roughly tripled since the end of 2025, exceeding $30 billion（約4.8兆円） in early 2026.

Why it matters

Anthropic's scale-up is directly translating into revenue for the chip makers and cloud providers that power Claude. Amazon's chip business exceeded $20 billion（約3.2兆円） in annualized revenue in the first quarter, growing at triple-digit rates. Google Cloud revenue skyrocketed 63% year over year last quarter, with a backlog that doubled to $460 billion（約74兆円）. Broadcom's AI chip revenue more than doubled in the most recent quarter, and Nvidia's data center revenue surged 85% year over year. These are the companies positioned to capture value from the AI infrastructure buildout.

What to watch

Broadcom expects AI chip revenue to exceed $100 billion（約16兆円） in fiscal 2027, more than double its trailing-12-month total revenue of $75 billion（約12兆円）. Anthropic will gain access to multiple gigawatts of next-generation TPUs starting in 2027. Nvidia is set to ramp up its next-generation Vera Rubin platform later this year, designed for agentic AI workloads.