Anthropic's annualized revenue has crossed $30 billion, according to a person familiar with the figures, pulling ahead of OpenAI's most recently disclosed run-rate of roughly $25 billion. The reversal, first reported earlier this month, is the cleanest quantitative signal yet that the market for frontier language models is no longer a single-company story.
The more striking number is efficiency. Anthropic's training compute spend in 2025 was approximately one-fourth of OpenAI's over the same period, according to two people with direct knowledge. That delta — roughly similar revenue at a fraction of the training bill — has begun to color conversations in OpenAI's investor base.
A partner at one of OpenAI's largest institutional backers described the mood this week as 'nervous but not panicked.' OpenAI's consumer franchise, the firm argues, is a defensible moat that revenue alone doesn't capture. But several of OpenAI's tender investors are now quietly asking for more detail about unit economics before the next secondary window opens.
Key facts
- 01Anthropic. A key thread of reporting in this story.
- 02OpenAI. A key thread of reporting in this story.
- 03Revenue. A key thread of reporting in this story.
Anthropic's shift toward per-token billing — away from the flat enterprise deals that dominated its 2024 book — has also mattered. The company now collects revenue that tracks usage almost linearly, which has made its growth line easier for auditors and sophisticated buyers to underwrite.
Both companies are widely expected to file for public offerings within the next eighteen months. The filings will be the first apples-to-apples comparison the market has ever gotten of foundation-model economics.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
