Amazon and Anthropic have expanded their infrastructure partnership into a much larger long-term commercial deal, with both companies now tying capital, cloud demand, and chip capacity together.

What changed

Amazon said it will invest $5 billion in Anthropic now and may invest up to another $20 billion later if certain commercial milestones are met. Separately, both companies said Anthropic has agreed to spend more than $100 billion over the next ten years on AWS technologies.

According to the two companies' announcements, the agreement secures Anthropic access to up to 5 gigawatts of compute capacity for training and serving Claude. That includes current and future generations of Amazon's custom AI hardware, spanning Trainium2 through Trainium4, along with Graviton CPU capacity. The deal also brings the full Claude Platform directly into AWS accounts, alongside Claude's existing availability through Amazon Bedrock.

Why it matters

The most important part is not the financing alone. It is that Anthropic is locking in a decade-long infrastructure commitment at a time when frontier model labs are competing for scarce power, chips, and data center buildouts.

For AWS, the arrangement is another sign that custom silicon is becoming a strategic lever, not just a cheaper alternative to Nvidia systems. For Anthropic, it is a concrete way to secure more capacity and broader inference deployment in Asia and Europe without changing its primary cloud provider. The conservative takeaway is that frontier AI partnerships are starting to look less like ordinary cloud contracts and more like industrial supply agreements.