Anthropic Surpasses OpenAI in Revenue: A Shift Redrawing the Commercial Balance of Power
Anthropic surpassed its rival OpenAI in annualized run-rate for the first time, thanks to its focus on enterprises rather than consumers. The numbers, the secret of the difference, the profitability angle, and a balanced reading.
A year ago, the consensus was that OpenAI's lead was insurmountable: the most famous brand, the largest user base, and the head start. But one number overturned this consensus in 2026. Anthropic, the maker of Claude, surpassed its rival OpenAI in "annualized run-rate revenue," in a shift that redraws the commercial balance of power in the AI industry. More important than the number itself is what stands behind it: two radically different business models testing a fundamental question — does profitability come from consumer scale or enterprise discipline?
The Numbers: A Story of Unprecedented Acceleration
The decisive shift happened in April 2026, when Anthropic's annualized run-rate surpassed OpenAI's for the first time. The growth curve is striking: from about one billion dollars in early 2025, to 9 billion by year-end, then 14 in February 2026, about 30 in April, and reaching 47 billion by May per CNBC reports. For comparison, Salesforce took about twenty years to reach 30 billion; Anthropic did it in under three years from a standing start. OpenAI's disclosed figure ranges between 25 and 33 billion. The overtaking was reported in Fortune's coverage on July 2, 2026.
The Secret of the Difference: Enterprise vs. Consumers
The story is not in the number but in its structure. About 80% of Anthropic's revenue comes from enterprise and developer customers via pay-per-token programmatic access. By contrast, OpenAI relies about 85% on ChatGPT consumer subscriptions, although about 95% of its users pay nothing. This difference is economically decisive: an enterprise customer generates three to five times more revenue per token than a consumer, its query patterns are more regular and cheaper to serve, and its contracts are "sticky," hard to abandon once Claude enters a critical business process.
The Depth of Enterprise Penetration
The detailed numbers reveal the depth of the shift. The number of Anthropic customers spending more than a million dollars annually exceeded a thousand by April, after being 500 less than two months earlier — doubling in record time. And eight of the Fortune 10 became customers. Most importantly, an independent signal from the payments company Ramp: it recorded Anthropic's share of corporate AI spending at 34.4% versus 32.3% for OpenAI in May — the first reversal of its kind. But the competition is not zero-sum: 79% of OpenAI's enterprise customers also pay for Anthropic.
The Deeper Dimension: Profitability and Training Cost
What makes the difference structural rather than temporary is the cost side. Per Wall Street Journal projections, OpenAI is heading toward spending about 125 billion dollars a year on training by 2030, while Anthropic estimates its spending for the same period at about 30 billion — a gap of nearly four times. The result is that Anthropic reported profitability in the last quarter of 2025 and expects positive free cash flow by 2027, while OpenAI projects a 14-billion-dollar loss in 2026 with no profitability before 2029 or 2030. Analyses say Anthropic's trajectory is closer to enterprise software economics (like Salesforce) than to infrastructure economics.
A Balanced Reading: What the Number Does Not Say
Despite the news's importance, it should be read carefully, as the sources themselves cautioned. First, "run-rate" is not audited annual revenue; it is a current pace multiplied by twelve, not a confirmed accounting figure. That is, the April overtaking does not prove Anthropic actually generated greater full-year revenue. Second, the figures on both sides are self-reported by the two companies, not by an independent party. Third, OpenAI still has the cleaner multi-year growth story, and a massive consumer base (more than 900 million weekly users) that may convert into enterprise revenue if its bet on agents succeeds. So the picture is a real shift, not a "final verdict."
This development remains a sample of a deeper shift in the industry: the end of the "expansion at any cost" era in favor of a discipline where buyers demand measurable returns. As both companies head toward an anticipated IPO, the S-1 filings will put real numbers before investors, making this divergence between the two business models impossible to ignore. For the developer and the enterprise, the practical lesson is that competition is no longer over "the smartest model" alone, but over who builds a sustainable economy around it — which serves whoever builds on a solid foundation rather than a deferred promise. (Disclosure: this news concerns Anthropic, the maker of Claude, and we have taken care to present it objectively and in balance.)