AMAZON DEEPENS ANTHROPIC AI INFRASTRUCTURE BET

Amazon is preparing to invest up to another $25 billion in Anthropic, deepening one of the most important partnerships in the artificial intelligence sector as demand for Claude continues to strain the startup's infrastructure.

- 

The expanded agreement gives Anthropic access to up to 5 gigawatts of Amazon compute capacity and commits the AI company to spending more than $100 billion on Amazon Web Services technologies over the next decade.

The deal adds another layer to a relationship that began in 2023 and has already made Amazon one of Anthropic's most important strategic backers. Amazon previously invested $8 billion in the Claude developer. Under the new agreement, Amazon will provide $5 billion immediately, with up to $20 billion more tied to commercial milestones. Anthropic, in return, is committing to AWS infrastructure at a scale that makes the pact as much a cloud and chip agreement as a conventional investment.

Anthropic said the partnership is designed to ease the “inevitable strain” caused by rising demand for Claude. That phrase captures the new reality of the AI market. The bottleneck is no longer only model quality or enterprise interest. It is physical infrastructure: chips, power, data centers, cooling, and the ability to serve millions of queries reliably as AI moves deeper into business workflows.

AMAZON TIES CAPITAL TO COMPUTE DEMAND

The structure of the agreement shows how AI investment has changed. Amazon is not simply writing a large check for equity exposure. It is linking investment, cloud demand, and custom chip adoption into one long-term commercial relationship.

Anthropic will spend more than $100 billion over 10 years on AWS technologies, with a major focus on Trainium, Amazon’s custom AI chip family. Amazon says the agreement covers current and future generations of Trainium chips, along with tens of millions of Graviton CPU cores. That means the partnership is designed to strengthen AWS’ hardware position as much as Anthropic’s model roadmap.