Anthropic signs Google and Broadcom deal for multi-gigawatt TPU capacity starting in 2027

The announcement

Anthropic said on April 7, 2026, in an X post and a matching company announcement, that it has signed a new agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity. The company expects that compute to begin coming online in 2027, where it will be used to train and serve frontier Claude models. By Anthropic’s own description, this is its largest compute commitment so far.

The company paired the infrastructure update with unusually direct commercial metrics. Anthropic said run-rate revenue has now exceeded 30 billion dollars, up from roughly 9 billion dollars at the end of 2025. It also said the number of business customers spending more than 1 million dollars on an annualized basis has risen from more than 500 in February to more than 1,000 today. That makes the TPU deal look like a response to real demand, not just a speculative capacity grab.

Anthropic also stressed that it runs Claude across AWS Trainium, Google TPUs, and NVIDIA GPUs, with most of the new compute to be located in the United States. Amazon remains its primary cloud provider and training partner, while Claude is still distributed across Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry. The broader signal is that frontier AI competition is increasingly about durable access to chips, power, and cloud infrastructure as much as it is about model quality.