Anthropic’s second-quarter revenue reached roughly $11.5 billion to $12 billion, according to a Wall Street Journal review of financial documents, surpassing OpenAI’s $6.7 billion in the same period for the first time. The numbers upend a hierarchy that had held since the modern AI boom began: OpenAI, the company that launched ChatGPT and defined the category, is no longer the revenue leader among the private AI labs.
The documents show OpenAI’s revenue grew 18% from the prior quarter, a respectable pace for most companies and a disappointment for some of its investors, who have grown used to steeper curves. Its losses, meanwhile, widened. OpenAI recorded a quarterly loss of $12.3 billion, the documents show, even as Anthropic pushed its operating result into positive territory.
The revenue comparison flatters Anthropic in one important way: the two companies count money differently. Anthropic’s figure includes substantial compute partnerships and infrastructure deals, while OpenAI’s reported number is closer to pure product revenue. Analysts said the gap is real but narrower than the headline suggests, and that both companies are selling essentially the same thing: access to frontier AI, delivered either as an enterprise product or a consumer subscription.
Even so, the reversal carries weight. Anthropic, which has positioned itself as the safety-first lab and the enterprise favorite, has been closing the gap for a year. Its annualized revenue had already been reported by market watchers at more than $65 billion on a run-rate basis. The formal quarterly figure, published this week, converts that speculation into a statement of fact: the challenger has caught the incumbent.
The financial shift is happening as both companies spend at records. OpenAI’s $12.3 billion quarterly loss is a measure of how much it costs to run, and to keep improving, the most-used AI products in the world. Anthropic’s positive operating result does not mean it is cheap to run; it means its revenue has finally outrun its costs at the operating line, a point the company has been chasing since its founding. Behind both numbers is the same engine of expense: computing power, bought by the megawatt and paid for by the month, that gets more expensive as models get bigger.
The money is moving in both directions. On Wednesday, Bloomberg reported that a Texas data center tied to Anthropic signed a $1.3 billion private-credit loan, one of the largest financing deals yet for AI infrastructure backed by a model lab. The loan will fund the computing capacity Anthropic needs to train and run its next models, the same buildout that has pushed cloud giants and data-center landlords to borrow at unprecedented scale. Private-credit lenders, once a niche corner of finance, have become the default bankers for the AI buildout, and deals of this size are now routine.
The backdrop is OpenAI’s preparations for a long-anticipated initial public offering. The company has been positioning itself for public markets for months, and the revenue slowdown, relative to its rival, arrives at an awkward moment. Bankers and analysts said OpenAI’s IPO valuation will ultimately be set by its own growth story, but the comparison with Anthropic gives investors a yardstick they did not have before. The two companies will also be selling the same story to the same buyers: that frontier AI is a durable, expanding market, not a fad with a short half-life.
For the broader industry, the reversal is a sign that the AI market has room for more than one winner. OpenAI defined the consumer moment with ChatGPT, and its brand remains the strongest in the category. Anthropic has built its franchise on enterprises, on coding assistants and on contracts with businesses that want AI without the public-relations risk. The two strategies were always going to meet somewhere; they have met in the revenue column.
What happens next depends on whether Anthropic can hold its lead. Its enterprise momentum is real, analysts said, but OpenAI’s consumer base is enormous, and its next model cycle could reset expectations. The burn rates on both sides are also climbing, which means the gap in the quarter could narrow or widen quickly depending on who spends first and who spends better. Both companies are racing to raise the capital that will fund the next generation of models, and the cost of compute continues to climb with every doubling of scale.
For now, the scoreboard has changed. The company that was supposed to be the cautious alternative has become the one with the bigger top line, and the company that invented the category is the one rushing to explain its slower quarter to investors. The AI race, by the numbers, has a new leader, and the gap between the two companies will be measured in quarters, not years. Both sides are spending as if the next model decides everything, because in this market, it might.
