Anthropic and Amazon Are Going All In on Compute: $100 Billion and 5 Gigawatts

Anthropic just announced a massive expansion of its partnership with Amazon, and I mean massive. We’re talking a commitment of over $100 billion over the next decade for AWS technologies, securing up to 5 gigawatts (GW) of new compute capacity to train and run Claude. To put that in perspective, 5 GW is enough to power several medium-sized cities. This isn’t just a press release; it’s a signal that Anthropic is betting the farm on Amazon’s custom silicon.

The agreement covers Graviton and Trainium chips from the current Trainium2 all the way through Trainium4, with options for future generations. Significant Trainium2 capacity is already coming online in Q2 2026, and scaled Trainium3 capacity is expected later this year. Nearly 1 GW of combined Trainium2 and Trainium3 capacity will be live by the end of 2026. That’s a lot of chips. Last I checked, they were already using over a million Trainium2 chips for training and serving Claude, and this deal will expand that dramatically.

This isn’t their first rodeo together. Anthropic has been working with Amazon since 2023, and over 100,000 customers now run Claude on Amazon Bedrock. They also launched Project Rainier together, one of the world’s largest compute clusters. But this new agreement goes further in three key ways.

Infrastructure at Scale

The $100 billion commitment isn’t just about raw capacity. It’s about locking in access to Amazon’s custom silicon roadmap. Andy Jassy, Amazon’s CEO, is clearly proud of what they’ve built: “Our custom AI silicon offers high performance at significantly lower cost for customers, which is why it’s in such hot demand.” He’s not wrong. Trainium chips are designed specifically for AI workloads, and they’re cheaper than NVIDIA’s alternatives. That matters when you’re spending nine figures a year on compute.

Claude Platform on AWS

This is the part that caught my attention. The full Claude Platform will be available directly within AWS. Same account, same controls, same billing. No extra credentials, no new contracts. For enterprises, this is huge. It means they can access Claude’s full feature set without jumping through compliance hoops. Claude remains the only frontier AI model available on all three major cloud platforms: AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry. That’s a strategic flex, but the deep integration with AWS is where the real value lives.

Continued Investment

Amazon is also putting more money where its mouth is: $5 billion today, with the potential for up to an additional $20 billion in the future. That’s on top of the $8 billion Amazon has already invested. Dario Amodei, Anthropic’s CEO and co-founder, summed it up well: “Our users tell us Claude is increasingly essential to how they work, and we need to build the infrastructure to keep pace with rapidly growing demand.”

Meeting Record Demand

Here’s where things get real. Enterprise and developer demand for Claude has accelerated in 2026, but the real story is consumer usage. Anthropic’s run-rate revenue has surged to over $30 billion, up from about $9 billion at the end of 2025. That’s tripling in less than a year. But growth at this pace puts strain on infrastructure. The company admits that “unprecedented consumer growth, in particular, has impacted reliability and performance for free, Pro, Max, and Team users, especially during peak hours.” Translation: Claude is getting hammered, and it shows.

This new capacity should help. They’re promising meaningful compute in the next three months, with nearly 1 GW total by year-end. Combined with a diversified hardware strategy—spreading workloads across different chips—Anthropic is trying to build the infrastructure to keep Claude at the frontier. Whether that’s enough to keep up with demand remains to be seen, but it’s a hell of a bet.

For more details, check out Anthropic’s official announcement. Or just ask Claude about it—if the servers aren’t overloaded.