Anthropic has committed to roughly 3.5 gigawatts of Google Cloud TPU capacity through 2031, with Broadcom involved in the buildout -- one of the largest disclosed compute contracts in the history of AI. The deal deepens Anthropic's alliance with Google and signals a deliberate diversification away from Nvidia GPUs at the very top of the market, where capacity, not cleverness, increasingly decides who can train and serve frontier models.
The compute news lands alongside a striking revenue milestone. Anthropic's annualized run rate has reportedly reached about $30 billion, up from roughly $14 billion in February, driven by enterprise API usage and its partnerships with Amazon and Google. That figure would put Anthropic ahead of OpenAI on revenue, a reordering of the perceived hierarchy that would have seemed implausible a year ago.
“Anthropic pairing record revenue with a 2031 compute reservation is a statement that it intends to be a structural winner, not just a fast follower.”
Together, the two data points reframe the race. Frontier AI is becoming a long-dated capacity-and-balance-sheet contest: the labs that can lock gigawatts of compute years in advance and fund it with real enterprise revenue are pulling away from everyone else. Anthropic pairing record revenue with a 2031 compute reservation is a statement that it intends to be a structural winner, not just a fast follower.