Amazon pledges up to $25 billion more to Anthropic, deepening AI cloud race

Amazon said it will invest up to $25 billion more in Anthropic, a deal that puts cloud capacity and custom chips at the center of the AI race. The package includes $5 billion committed immediately and up to $20 billion more if commercial milestones are met, while Anthropic said it expects to spend more than $100 billion over the next 10 years on Amazon cloud technologies.

The scale matters because Anthropic is not just buying access to more servers. It said the expanded agreement could secure up to 5 gigawatts of compute for Claude, with nearly 1 gigawatt of Trainium2 and Trainium3 capacity expected online by the end of 2026. Amazon said the arrangement leans on current and future generations of Trainium chips and tens of millions of Graviton cores, a reminder that the fight for AI market share is now as much about infrastructure economics as model quality.

For Amazon, the deal extends a relationship that began in September 2023, when it said it would invest up to $4 billion and AWS became Anthropic’s primary cloud provider for mission-critical workloads. In November 2024, Amazon added another $4 billion, bringing its disclosed investment to $8 billion before this latest expansion. Amazon also said more than 100,000 customers run Claude models on AWS, and that Claude is one of the most popular model families on Amazon Bedrock. Anthropic said the new capacity also includes inference infrastructure in Asia and Europe, a sign it is building for international demand, not just U.S. enterprise buyers.

That shift has direct implications for monday.com, where AI is increasingly part of the product story rather than a side experiment. The company said in 2025 that its AI strategy centers on AI Blocks, Product Power-ups and the Digital Workforce. It also said users completed more than 26 million AI-driven actions across the platform by the end of the first quarter of 2025, with adoption up more than 150% quarter over quarter. In 2026, monday.com said external AI agents can sign up, authenticate and operate directly within the platform alongside human teams.

For product and engineering teams, the message is clear: buyers will judge work software not only on interface design, but on model access, reliability and cost. If foundation models become more expensive to run at scale, embedded AI features will face tougher margin math. Sales teams will also have to answer harder questions from customers asking whether a work platform can keep AI responsive for long-running agents, context-heavy workflows and global deployments without leaning too heavily on a small set of cloud-model alliances.