Anthropic Overtakes OpenAI on Revenue, and Claude Code Is Why

Fortune reported on July 2 that Anthropic's revenue has overtaken OpenAI's. Anthropic is running at a $47 billion annualized rate as of May, against OpenAI's self-reported $25 to $33 billion. Eighteen months ago that gap ran the other way, by a wide margin.

The interesting part isn't the crossover. It's why it happened, and the answer is more structural than a single product win.

Two different businesses wearing the same "AI lab" label

OpenAI's revenue is mostly consumer subscriptions: ChatGPT Plus, Pro, and a long tail of API usage. Anthropic's is mostly enterprise contracts: large, recurring deals for customer service automation, content analysis, and code generation, sold directly to companies that sign multi-year agreements.

Subscription revenue churns. Enterprise contract revenue, once it's live inside a company's workflow, doesn't get ripped out casually. That difference alone explains a lot of the durability gap between the two numbers, independent of which lab has the smarter model this quarter.

Claude Code is the single biggest line item

Anthropic's fastest-growing product isn't a chat assistant, it's a coding agent. Claude Code's annualized revenue hit $2.5 billion by February 2026, and it's continued climbing since. That's not a side project generating rounding-error revenue. It's a meaningful chunk of a $47 billion run rate, from a product that didn't exist in its current form two years earlier.

Developer tools turning into a lab's largest revenue driver is the part worth sitting with. The market that decides whether Anthropic or OpenAI wins isn't the general chatbot market anymore. It's increasingly the market for tools that write and ship code, which is exactly the audience reading this sentence.

The valuation followed the revenue

Anthropic closed a $65 billion Series H at a $965 billion valuation in late May, led by Altimeter, Dragoneer, Greenoaks, and Sequoia. That figure put Anthropic's valuation above OpenAI's for the first time, a genuine reversal from a company that spent its first several years as the clear number two.

[!NOTE] Self-reported revenue figures from both labs should be read with the usual caution; neither company is a public filer with audited quarterly numbers. The consistent theme across independent coverage, though, is the direction of the trend, not just the raw figures, and that direction has been stable for several quarters now.

Where this leaves the model race

The competitive pressure is showing up everywhere at once this month. Anthropic restored Claude Fable 5 and Mythos 5 access globally on July 1 after US export controls affecting them were lifted, adding cybersecurity classifiers it's calling its strongest safeguards yet. On the benchmark side, Opus 4.8 scores 69.2% on SWE-Bench Pro, with Sonnet 5 at 63.2%, both meaningfully ahead of the field on agentic coding tasks specifically.

Meanwhile OpenAI answered by merging ChatGPT and Codex into a single agent the following week, a release explicitly framed by its own coverage as a response to intensifying Anthropic competition.

Revenue crossovers in fast-moving markets aren't permanent. But this one reflects something structural: Anthropic bet on developers and enterprises early, that bet is now the majority of its book, and the rest of the industry is visibly reorganizing product strategy around the same bet.