anthropic reportedly passed openai in revenue. here is what actually flipped

the reported numbers

Industry reporting this month puts Anthropic at roughly $47 billion in annualized revenue as of May, with the company said to be profitable this year, while OpenAI's annualized revenue is reported in the $25 to 33 billion range. Corporate card data from Ramp showed Anthropic overtaking OpenAI in business subscriptions back in May. And Claude Code, a developer tool, reportedly reached $2.5 billion in annualized revenue by February.

The usual caveats apply: these are reported figures from private companies, annualized run rates rather than audited results, and the numbers move monthly. But even with generous error bars, the direction is unambiguous. The company with a fraction of the consumer fame built the bigger business.

how the quiet one won

OpenAI won the consumer story. ChatGPT is the household name, the app your aunt knows. That was supposed to be the moat.

Meanwhile, Anthropic did something less glamorous: it went where the money actually was. Coding tools. Agents that do work. Enterprise deployments. Business subscriptions. The unsexy pattern held: consumers pay a little sometimes, businesses pay a lot on a contract.

The Claude Code number is the tell. A terminal tool for developers, no consumer appeal whatsoever, reportedly worth $2.5 billion a year on its own. Compare the energy the industry spent on viral image generators and companion apps, and then look at where the revenue showed up.

the lesson that transfers

You are probably not running a frontier lab, but the mechanics scale down perfectly:

• Fame is not revenue. The most-known product and the best business are often different companies. Followers, downloads, and press are inputs, not the score.
• Sell into workflows, not moments. ChatGPT usage is moments: a question, an essay, a laugh. Claude's growth is workflows: code shipped, documents processed, agents running all day. Workflows renew. Moments churn.
• Boring buyers are the best buyers. A business that saves 10 hours a week does not cancel in month two because the novelty faded.

I build for creators, and I preach the same thing to them: the audience is the top of the funnel, not the business. The business is whatever people pay you for repeatedly, and it is usually less glamorous than the content.

what to watch next

If these numbers hold, every lab now has proof that enterprise agents and developer tools are where the market pays. Expect the consumer-facing free tiers to keep getting shockingly good, they are marketing budget now, while the real product race happens in tools that do work.

Which, if you are a builder on a budget, is the best possible outcome: the giants subsidize your tooling while they fight over enterprises. Take the free leverage and build something boring that people pay for. It apparently works.