Anthropic hits $9B revenue run rate in 2025, up from $4B in July, as new funding round takes shape
Jan 22, 2026
Key Points
- Anthropic's annualized revenue run rate hits $9B in 2025, up from $4B in July, falling short of the 10x growth trajectory CEO Dario Amodei had previously suggested.
- Inference costs ran 23% higher than projected due to prioritizing service availability over unit economics, pressuring gross margins below the company's 40% guidance.
- Anthropic is moving to optimize margins by purchasing TPUs directly, building new data centers, and routing workloads across cheaper legacy models and specialized silicon.
Summary
Anthropic's revenue run rate reached $9B in 2025, up from $4B in July. That represents a 125% increase over five months, below the 10x growth trajectory some had anticipated. CEO Dario Amodei had suggested the company might repeat its prior jump from $100M to $1B, but cautioned that sustaining that pace would be "pretty crazy" at this scale. At $9B annualized, Anthropic remains two orders of magnitude below the $1 trillion annual revenue baseline that only a handful of companies globally reach.
Lightspeed Venture Partners and Menlo Ventures are set to co-lead a new funding round. The segment did not disclose valuation or capital amount.
Gross margin pressure exists but is manageable. The Information reported that Anthropic's inference costs on Google and Amazon servers ran 23% higher than projected. The company had prioritized service availability during explosive growth over unit economics. Anthropic previously guided to 40% gross margins, but actual results came in lower.
The company is pursuing multiple cost levers. It is now buying TPUs in addition to relying on cloud providers, planning new data center builds, and expects to benefit from load balancing inference across heterogeneous silicon. Basic workloads will route to cheaper legacy models while inference-hungry tasks go to specialized stacks like Cerebras or Grok. Whether Anthropic can sustain pricing power depends on whether its output delivers enough value to keep customers willing to pay more rather than optimize toward lower cost-per-token.