Amazon Invests $25 Billion in Anthropic, Securing 5 Gigawatts of Compute for Claude
The deal includes $100B in AWS spending and over 1 million custom Trainium chips to power Claude.
Amazon has made a monumental $25 billion equity investment in AI lab Anthropic, the largest deal of its kind in the cloud AI market. The agreement secures a staggering 5 gigawatts of compute capacity on Amazon Web Services (AWS) dedicated to training and serving Anthropic's Claude models. This power allocation is equivalent to running five million American homes. The partnership also locks in over $100 billion in AWS spending over the next decade and involves more than one million of Amazon's custom Trainium AI accelerator chips already in operation.
The strategic heart of the deal is Project Rainier, a super-cluster built specifically for Anthropic containing nearly 500,000 Trainium2 chips. This massive investment in custom silicon is Amazon's play to reduce dependence on Nvidia and control AI infrastructure costs, similar to Apple's shift to M-series chips. Anthropic becomes the definitive anchor customer, proving Trainium can handle frontier model workloads. The move follows Amazon's recent $50 billion deal with OpenAI, signaling AWS's intent to host all major AI players and not be locked into a single vendor, while simultaneously guaranteeing Anthropic's compute future amid intensified competition.
- $25 billion equity investment with an additional $100+ billion committed to AWS spending over 10 years.
- Secures 5 gigawatts of compute power—equivalent to five million homes—for Claude's training and deployment.
- Anchors Amazon's custom silicon strategy with the 500,000-chip 'Project Rainier' cluster running on Trainium processors.
Why It Matters
This reshapes cloud AI economics, reduces Nvidia's grip, and guarantees compute for a leading AI model amid a capacity arms race.