Anthropic reveals $30bn run rate and plans to use 3.5GW of new Google AI chips

Deal essentials

Broadcom has filed that Google asked it to design and supply next‑generation TPUs and networking components for future AI racks — and, it has been reported that, Anthropic will access about 3.5 gigawatts of those TPU-based accelerators starting in 2027. The news comes from a Broadcom regulatory filing that also outlines a supply-assurance pact to cover networking and other parts for Google’s AI racks through 2031. Big silicon, big racks, big ambitions.

Money and caution

It has been reported that Anthropic also told markets its run‑rate revenue has surpassed $30 billion, up from about $9 billion at the end of 2025, and that more than 1,000 enterprise customers are now each spending over $1 million annually. That’s the emotional punch: dizzying growth, almost cinematic. But Broadcom’s filing tempers the fanfare — the deployment of 3.5GW is “dependent on Anthropic’s continued commercial success,” and the chipmaker says parties are negotiating with operational and financial partners. In short: scale is possible, but someone still has to underwrite it.

Why it matters

This is a slice of the larger trend: hyperscalers and chip partners carving up the future of AI infrastructure. Broadcom’s CEO Hock Tan has been vocal about third parties building custom accelerators for hyperscalers, and Google’s move to farm out next‑gen TPU design is another marker of that strategy. Anthropic isn’t putting all its eggs in one basket either — it reportedly uses Google Cloud, AWS Trainium and Nvidia gear to “match workloads to the chips best suited for them.” Can Anthropic sustain the revenues and the cash flow to fill those racks? That’s the question investors and partners now have to answer.

Sources: The Register