Anthropic signs an agreement with Google and Broadcom for multiple GWs of TPU capacity, and says run-rate revenue has crossed $30B, up from ~$9B at 2025's end
We have signed a new agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity that we expect to come online starting in 2027.
Context & Ripple Effects
This expands Anthropic’s earlier Google cloud partnership, which provided 1 million TPUs and 1 GW of capacity for 2026; the new agreement shifts the relationship toward a larger, longer-dated supply commitment for next-generation hardware (the initial 1 GW TPU arrangement).
The accompanying revenue disclosure gives the infrastructure order commercial context: Anthropic is pairing rapid reported demand growth with capacity intended to arrive from 2027, rather than treating compute as a short-term spot purchase.
First-order effects
- Anthropic gains a contracted path to multiple gigawatts of next-generation TPU capacity beginning in 2027, improving visibility into the compute it can plan models and customer growth around.
- Google secures a major TPU-cloud customer, while Broadcom receives a clearer demand signal for future Google TPU production; related filings describe an expanded arrangement of about 3.5 GW (the reported expansion of Anthropic’s capacity access).
Second-order effects
- The deal makes TPU capacity a more credible alternative for frontier-model workloads, pressuring competing cloud and chip providers to offer similarly dependable, multi-year capacity commitments rather than only flexible on-demand access.
- Longer commitments concentrate planning across the cloud operator, chip designer, and customer: hardware roadmaps, data-center buildouts, and Anthropic’s deployment schedule become more interdependent.
Third-order effects
- If such contracts become standard, frontier AI infrastructure will increasingly be allocated through long-duration bilateral agreements, favoring companies able to finance capacity well before it is operational.
- The industry could shift from generalized cloud competition toward vertically coordinated compute supply chains, where model developers’ growth plans help determine custom-chip production and data-center investment.
The trend: This is a data point in the shift from opportunistic AI compute procurement to long-term, utility-scale capacity contracting tied to custom silicon.