Anthropic told investors its annualized revenue has reached $65 billion, up from roughly $1 billion two years ago, and the company is now in talks that could value it near $2 trillion when it eventually goes public. Annualized revenue run rate means taking the most recent month or quarter of sales and multiplying it out to a full year, so it is a snapshot of current pace, not booked cash. The number is being driven by enterprise API contracts, the deals where a bank or an insurer wires Claude into its own claims system or trading desk. That is the deployment side of the business: not how good the model scores on a benchmark, but how many paying seats and production integrations a lab has actually locked in. Anthropic's compute partner is SpaceX, whose Colossus cluster in Memphis supplies the Nvidia GPU capacity Anthropic trains its next model generation on.

The same week, Ars Technica reported Anthropic and OpenAI are now cutting API prices in response to Chinese labs like DeepSeek and Alibaba's Qwen team, whose open-weight models are cheaper to run and increasingly close on coding and reasoning evals. That is the constraint the $2 trillion number sits inside: Anthropic's revenue is compounding off contracts signed before the Chinese price war intensified, and every contract renewal from here has to hold against a rival model a customer can download once and run at a fraction of the API cost. The next test is Anthropic's next enterprise renewal cycle, where procurement teams now have that comparison in hand before they sign.