What Happened

For most of the generative-AI boom, OpenAI was the company everyone measured themselves against. That framing broke this summer. In a July 2, 2026 analysis, Fortune confirmed what a string of data providers had been signaling for weeks: Anthropic, the maker of Claude, has overtaken OpenAI on revenue — and the gap is widening rather than closing.

The reversal is striking because the two labs are chasing the same technology with almost opposite go-to-market strategies. Anthropic has bet on developers and enterprises consuming Claude through paid APIs. OpenAI built the most famous consumer product in AI history in ChatGPT. In 2026, the API bet is winning the revenue race — and the divergence is now visible in the raw financial numbers, not just in narrative.

The Numbers

The headline figures tell the story on their own. As of mid-May, Anthropic was tracking toward roughly $47 billion in annualized (run-rate) revenue, according to figures cited by Fortune, while OpenAI's most recent disclosures pointed to $25–33 billion for 2026. Analysts tracking the two companies say Anthropic's run rate has since pushed close to double OpenAI's, with the business reportedly adding on the order of tens of millions of dollars in annualized revenue every single day.

The valuation picture moved in the same direction. In late May, Anthropic closed a mega-round that pushed its valuation toward the $1 trillion mark, surpassing OpenAI's for the first time and making it, on paper, the most valuable AI startup in the world. On profitability, Anthropic has told investors it expects to be in the black by 2029 — a year ahead of OpenAI's own timeline.

Two independent datasets underpin the shift. Enterprise-spend tracker Ramp found that Anthropic overtook OpenAI in business subscriptions back in May. And Similarweb data showed monthly visits to ChatGPT slipping below a majority of the generative-AI market for the first time that same month — a sign, as Deutsche Bank's Adrian Cox put it, that "consumers are increasingly willing to switch between models."

Two Opposite Business Models

The clearest explanation for the gap is structural, not a matter of one model simply being smarter than the other. The two companies make money in nearly mirror-image ways.

• Anthropic — API-first: roughly 75–85% of its revenue comes from usage-based APIs, where businesses and developers pay per token as they build Claude into their own products, agents, and internal tools. Usage scales with how much work customers push through the model.
• OpenAI — consumer-first: a large majority of its revenue — reported at anywhere from about 65% up to 85%, depending on the estimate — still comes from ChatGPT subscriptions, dominated by individual and Plus/Team users.

Those two engines grow very differently. A consumer subscription caps out at a flat monthly fee per user. An enterprise that wires Claude into a coding workflow or a customer-support pipeline can multiply its spend as usage climbs, which is why Anthropic has reported eye-watering net revenue retention — existing customers expanding their spend several times over year on year. Consumer subscriptions are stickier and more predictable; API usage is lumpier but scales far faster when adoption takes off.

The Enterprise and Coding Engine

Nowhere is Anthropic's momentum clearer than in software development. Claude Code, the company's agentic coding tool, has grown into a multibillion-dollar line of business in its own right, with a run rate reported to exceed $2.5 billion after more than doubling since the start of 2026. By some estimates Claude now accounts for a majority of dedicated coding-model spend across the industry.

The enterprise base has thickened to match. Anthropic has said it now has more than 1,000 customers spending over $1 million a year, roughly double the number from February. Governments are buying in too: California recently announced what it called the largest U.S. state-government AI deployment to date, giving every state agency — and any city or county that opts in — access to Claude at a 50% discount.

The pattern is consistent: Anthropic keeps landing where the money compounds — inside the tools that companies pay for by the hour of machine work — while OpenAI's flagship remains a product people pay a fixed price to chat with.

OpenAI's Response

OpenAI is hardly standing still — it remains one of the most-used consumer products on the internet and continues to ship frontier models. But the competitive pressure is visible in how CEO Sam Altman has begun to frame the moment. Around the same time the revenue figures landed, Altman argued for a new, U.S.-led international framework to govern advanced AI, pointing to aviation safety and the International Atomic Energy Agency as templates for global standards and oversight.

The pitch reframes the race as one about safety, trust, and the rules of the road rather than pure revenue ranking — a natural move for a company that is still winning on consumer reach and mindshare even as a rival pulls ahead on the income statement. It also underscores that "who's ahead" in AI now depends heavily on which scoreboard you read: users, revenue, valuation, or capability benchmarks can each crown a different leader.

Why It Matters

The reordering matters beyond bragging rights. For years, the assumption was that whoever owned the consumer relationship would own the AI economy. Anthropic's rise suggests the opposite may be true, at least for now: the biggest, most durable dollars are flowing through APIs and into the software that businesses run, where AI is billed as infrastructure rather than sold as a subscription.

For companies choosing where to build, the takeaway is that the "leader" label is splitting into categories. If you are shipping developer tooling, agents, or high-volume automation, Anthropic's enterprise-and-coding gravity is now hard to ignore. If your priority is the broadest consumer reach and a familiar chat product, OpenAI still sits at the center of that world. And with Google's Gemini pushing hard on the same enterprise turf, the safest bet for buyers is to stay portable — the data on model-switching shows users and businesses alike are increasingly unwilling to be locked in.

None of this is settled. Revenue run rates can move quickly at this stage, valuations are private and contested, and a single strong model release can reshuffle the deck. But the through-line of 2026 is clear: the AI market has more than one winner, and the metric you pick decides who it is. Fortune's full analysis is available in its July 2 report.