Anthropic locks in massive TPU deal with Google and Broadcom

Anthropic announced late Monday, April 6, that it has signed a new agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity. The new infrastructure is expected to come online starting in 2027 and will support future Claude models.

This is more than a routine supplier update. Anthropic called it its biggest compute commitment so far and tied the deal directly to surging enterprise demand. The company says its revenue run rate has now passed $30 billion, up from roughly $9 billion at the end of 2025.

For CIOs, the message is clear. Competition for AI infrastructure is getting tighter, and the top model labs are locking in capacity years in advance. Anthropic’s continued use of Google TPUs, AWS Trainium, and Nvidia GPUs also underlines that frontier AI is becoming a multi-platform infrastructure game, not just a model race.

The announcement also strengthens Google Cloud’s role in the AI stack while highlighting Broadcom’s growing importance as a supplier of networking and custom silicon components for hyperscale AI systems. It is another signal that frontier AI is now shaped as much by compute and data center strategy as by model quality.