Anthropic said its annual revenue run rate has surpassed $30 billion, up from roughly $9 billion at the end of 2025 -- a more than 3x increase in about a year, driven by explosive enterprise adoption of Claude and surging API consumption from developers building on the platform. The disclosure came alongside an expansion of its compute partnership with Google and Broadcom to secure approximately 3.5 gigawatts of next-generation TPU-based capacity beginning in 2027, in addition to more than 1 gigawatt already coming online in 2026.
The scale is hard to overstate. One gigawatt of sustained AI compute is roughly equivalent to Anthropic's entire fleet at the start of 2026, so the new commitment represents something close to a 4.5x expansion of its compute base within 18 months. Broadcom will supply networking and components for Google's next-generation AI racks through 2031 and route Google-designed TPUs to Anthropic as part of the deal.
“Broadcom will supply networking and components for Google's next-generation AI racks through 2031 and route Google-designed TPUs to Anthropic as part of the deal.”
The strategic logic is that compute, not model architecture, is now the binding constraint on frontier labs. By going deep on Google's TPUs, Anthropic both guarantees supply and hedges away from the Nvidia scarcity squeezing the rest of the industry. For a company reportedly heading toward a public listing, demonstrating both a tripling revenue ramp and locked-in compute is exactly the story institutional investors will want to underwrite.