Anthropic expands Amazon partnership with 5GW compute, $100B commitment, big bet on Trainium chips

Anthropic and Amazon Web Services (AWS) have dramatically deepened their partnership with one of the largest infrastructure commitments in AI history. The expanded collaboration centers on constructing up to 5 gigawatts of new data center capacity dedicated to AI workloads, a scale comparable to powering millions of homes. This build-out is underpinned by a staggering financial commitment that could reach $100 billion over the next decade, signaling a long-term, capital-intensive bet on the future of large-scale AI.

The partnership is a major strategic win for Amazon's custom silicon division. A core element of the deal is Anthropic's agreement to use Amazon's proprietary Trainium and Inferentia chips for training its future foundation models, like the Claude series, and for running inference. This provides a flagship customer and validation for Amazon's chip designs against competitors like NVIDIA's GPUs. For Anthropic, it secures guaranteed, massive-scale compute—the most critical resource for AI labs—and deepens its integration with AWS as its primary cloud provider.

This move intensifies the cloud infrastructure wars, with Google (a previous Anthropic investor) and Microsoft/OpenAI watching closely. It provides Anthropic with the firepower needed to continue scaling its models in the race against OpenAI's GPT and Google's Gemini, while giving Amazon a powerful ally in the foundational model layer to compete with Azure's OpenAI partnership. The scale of the commitment highlights the enormous capital and energy requirements of the next phase of AI development.