Anthropic and Amazon Are Going All In on Compute: $100 Billion, 5 Gigawatts, and a Decade of Trainium

Anthropic just announced a major expansion of its partnership with Amazon, and I have to say, the numbers here are staggering. We’re talking up to 5 gigawatts of new compute capacity for training and deploying Claude, with a commitment of over $100 billion to AWS technologies over the next decade. That’s not a typo.

Let me break down what’s actually happening, because there’s a lot more here than just a press release.

The Compute Hunger is Real

Anthropic has been working with Amazon since 2023, and over 100,000 customers now run Claude on Amazon Bedrock. Together they launched Project Rainier, one of the largest compute clusters in the world, and they’re currently using over one million Trainium2 chips. But that’s apparently not enough.

The new agreement secures up to 5GW of capacity, with significant Trainium2 capacity coming online in Q2 2026 and scaled Trainium3 capacity expected later this year. By the end of 2026, they’ll have nearly 1GW total of Trainium2 and Trainium3 capacity online. The commitment spans Graviton and Trainium2 through Trainium4 chips, with an option to buy future generations of Amazon’s custom silicon.

This is higher than I expected, honestly. 5GW is an enormous amount of compute. To put it in perspective, that’s enough to power several small cities. It shows just how compute-hungry frontier AI training and inference really are.

The Money Talk

Amazon is investing $5 billion in Anthropic today, with up to an additional $20 billion in the future. This builds on the $8 billion Amazon has previously invested. So we’re looking at a potential total of $33 billion from Amazon alone, plus the $100 billion infrastructure commitment.

But here’s the part that caught my attention: Anthropic’s run-rate revenue has now surpassed $30 billion, up from approximately $9 billion at the end of 2025. That’s tripling in roughly a year. Enterprise and developer demand for Claude has accelerated in 2026, and consumer usage across free, Pro, and Max tiers has exploded.

The Reliability Problem They’re Actually Solving

Anthropic is refreshingly honest about one thing: “our unprecedented consumer growth, in particular, has impacted reliability and performance for free, Pro, Max, and Team users, especially during peak hours.” Translation: Claude has been getting slow and flaky when too many people try to use it at once.

This is the real reason for the deal. They need the capacity to handle the demand, and they need it fast. The agreement promises “meaningful compute in the next three months” and nearly 1GW total before the end of the year. Combined with their diversified hardware strategy (workloads spread across a range of chips), they’re trying to keep Claude reliable while growing.

Claude Platform on AWS

One interesting detail: the full Claude Platform will be available directly within AWS. Same account, same controls, same billing, with more Claude Platform features and no additional credentials or contracts necessary. This is a big deal for enterprises that want to use Claude but need to stay within their existing AWS governance and compliance frameworks.

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 advantage Anthropic is smart to maintain.

My Take

This partnership is massive, but it’s also a bet. Anthropic is essentially tying its future to Amazon’s custom silicon roadmap, from Trainium2 through Trainium4 and beyond. If Amazon delivers on performance and cost, this could be a winning strategy. If not, Anthropic has locked itself into a single hardware supplier for a decade.

The $100 billion figure sounds impressive, but it’s spread over ten years. That’s $10 billion per year, which is roughly what they’re spending now on compute. So this isn’t a sudden spending spree; it’s a long-term commitment that gives them predictable capacity.

What I find most interesting is the timing. The deal comes as Claude’s consumer growth is straining infrastructure, and Anthropic needs to fix reliability before it hurts their brand. The Q2 capacity will help, but I’ll be watching to see if they can actually deliver on the performance promises.

Overall, this is a smart move for both companies. Amazon gets a marquee customer for its custom silicon and a long-term revenue stream. Anthropic gets the compute it needs to stay competitive. Whether it’s enough to keep Claude at the frontier remains to be seen, but at least they won’t be bottlenecked by hardware.