Anthropic Secures 16GW of Compute Capacity
According to a report from The Information, Anthropic has signed new compute capacity agreements totaling at least 14.8 gigawatts (GW) over the past 11 months, with the capacity expected to be deployed across multiple years. Adding the 1–2GW already locked before October last year, the company now has approximately 16GW of committed compute capacity. Based on public contracts, the potential total cost could reach up to $517 billion, with most of the spending occurring over the next decade.
Expansion Driven by Claude Demand
The expansion is largely fueled by demand for Anthropic's flagship AI model, Claude. The company noted that its Claude Code and Cowork products grew beyond expectations this year, prompting a centralized effort to secure compute resources.
Multiple Partners Provide Compute
On the partnership side, Amazon committed up to 5GW under a new agreement, while Google and Broadcom together provide an additional 5GW. Microsoft and NVIDIA contribute up to another 1GW. Additionally, Anthropic leased the entire compute capacity of SpaceX's Colossus 1 data center, gaining over 220,000 NVIDIA GPUs, including H100, H200, and GB200 models. Many of these contracts extend beyond 2030.
Comparison with OpenAI's Plans
In contrast, OpenAI has told investors it aims to reach 30GW of compute by 2030 and expects cumulative spending of about $750 billion by then. It should be noted that The Information's estimate of $517 billion is the maximum potential cost under existing cloud, chip, and data center contracts; some capacity may be deferred or go unused, and certain contracts include early termination clauses.