Something extraordinary just happened and most people missed it. Anthropic's annualized revenue has hit $30 billion, up from $9 billion at the end of 2025. That is more than tripling in roughly four months. And it means Anthropic has now surpassed OpenAI's run-rate revenue of approximately $25 billion.

Read that again. The company founded in 2021 by ex-OpenAI researchers who left because they thought safety was not being taken seriously enough is now generating more revenue than OpenAI itself. The safety company just became the biggest AI company on Earth.

The Numbers Tell a Story OpenAI Does Not Want You to Hear

The composition of Anthropic's revenue is what separates it from every competitor. Approximately 80% comes from business customers. The number of customers spending over $1 million annually has doubled to more than 1,000, up from 500 just two months ago. Business subscriptions to Claude Code have quadrupled since the start of 2026.

New data from Ramp, the spend-management platform that tracks which AI tools U.S. businesses actually pay for, confirms the trend. In March, 30.6% of U.S. businesses paid for Anthropic's tools, up from 24.4% in February. That is a 25% jump in enterprise adoption in a single month. Meanwhile, OpenAI's enterprise share has been flat.

This is not a story about one company getting lucky. This is a structural shift. Enterprises are choosing Claude over ChatGPT because it is better at the work that actually matters: coding, analysis, and complex reasoning. Claude Code alone has become a verb in developer circles the way "Google it" became a verb for search.

The Curse of Being the Best

There is a catch, and it is a brutal one. Being the best model is destroying Anthropic's infrastructure. In March 2026 alone, the company experienced five major platform outages. Claude Code users reported burning through their 5-hour subscription windows in under 90 minutes. Rate limits are being hit so fast that paying customers are furious.

This is what one analyst calls the "Inference Trap": build the best model, users surge, inference compute explodes, and you are forced to either throttle users, raise prices, or cannibalize the compute you need to train the next model. OpenAI hit this wall during the Ghibli moment in March 2025 when ChatGPT gained 100 million signups in a week. Now it is Anthropic's turn.

And then there is Mythos. Anthropic's newest model is roughly five times more expensive to run than Opus 4.6. It scored 93.9% on SWE-bench Verified (versus Opus 4.6's 80.9%), and found thousands of zero-day vulnerabilities across every major operating system and browser. It is so powerful that Anthropic will not release it publicly, instead restricting access to 12 partner companies through Project Glasswing.

The Subsidized Era Is Over

Both Anthropic and OpenAI are heading toward IPOs, and that changes everything. As one industry analyst writes, the era of subsidized AI model usage is ending. When these companies answer to public shareholders instead of venture capitalists, the math has to work. That means higher prices, stricter rate limits, and the end of "all you can eat" subscription models.

Anthropic projects positive free cash flow by 2027 to 2028 while spending roughly four times less on training than OpenAI. That is the kind of efficiency that Wall Street loves. Meanwhile, OpenAI is projecting $85 billion in annual spending while its CEO and CFO cannot agree on strategy.

What This Means

Two years ago, Anthropic was the underdog. The safety-obsessed startup that could not compete with OpenAI's marketing machine and consumer brand. Today, it is the biggest AI company on the planet by revenue, the one central banks are convening emergency meetings about, and the one whose model governments cannot decide whether to ban or buy.

The lesson is simple. In AI, the best model wins. Not the best brand. Not the biggest marketing budget. Not the loudest CEO. The best model. Anthropic built the best model, and the market followed.

Sam Altman should be losing sleep. Not because Anthropic is catching up. Because it already passed him. And it is accelerating.