Anthropic, Google, and Broadcom are going all-in on next-gen compute

Anthropic just dropped some serious news: they’ve signed a new agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity, expected to come online starting in 2027. This is a massive bet on infrastructure to power their Claude models and keep up with what they describe as “extraordinary demand” from customers worldwide.

Krishna Rao, Anthropic’s CFO, framed it as a disciplined scaling play. “We are building the capacity necessary to serve the exponential growth we have seen in our customer base while also enabling Claude to define the frontier of AI development,” he said. That’s a lot of corporate speak for “we’re spending a ton of money because people can’t get enough of this thing.”

And the numbers back that up. Anthropic’s run-rate revenue has hit $30 billion—up from about $9 billion at the end of 2025. That’s a 3x jump in roughly four months. When they announced their Series G fundraising in February, they mentioned over 500 business customers spending over $1 million annually. That number has now doubled to over 1,000 in less than two months. That’s not just growth; that’s a hockey stick.

Most of this new compute capacity will be built in the United States, which expands on their November 2025 commitment to invest $50 billion in American computing infrastructure. It’s a smart political and practical move—keeping critical infrastructure domestic while also playing into the whole “AI sovereignty” narrative.

This isn’t Anthropic’s first rodeo with Google. They already announced increased TPU capacity last October, and this deepens that relationship. Broadcom is also in the mix, which makes sense given Broadcom’s role in custom chip design and networking for hyperscale data centers.

What I find interesting is Anthropic’s hardware diversity. They train and run Claude on AWS Trainium, Google TPUs, and NVIDIA GPUs. That’s three different platforms, which gives them flexibility to match workloads to the best chip for the job. It also means they’re not locked into any single vendor’s roadmap—a smart hedge in a market where chip supply is notoriously volatile.

Amazon remains their primary cloud provider and training partner, with Project Rainier still in play. And Claude is still 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 unique position, and one that gives enterprise customers options.

All this compute doesn’t come cheap, and the environmental footprint is going to be significant. Gigawatts of capacity means a lot of power, a lot of cooling, and a lot of water. Anthropic hasn’t said much about sustainability here, but that’s going to be a growing conversation as these deployments scale.

Still, the message is clear: Anthropic is betting that the demand for frontier AI isn’t slowing down anytime soon. And with revenue growing this fast, they can afford to think big.