Anthropic posts over $11.5 billion in Q2 revenue and its first operating profit, narrowing the gap with OpenAI

Anthropic has told investors that its preliminary revenue for the second quarter of 2026 exceeded $11.5 billion, more than double the $4.73 billion it recorded in the first quarter, and that it reached positive adjusted operating income for the period. The figures, first reported by Bloomberg on August 15, mark a turning point for a company that only two years ago was widely viewed as the underdog in the race to build frontier artificial intelligence.

The scale of the acceleration is unusual even by the standards of the current AI boom. In the same quarter a year earlier, Anthropic booked just $787 million in revenue, meaning sales jumped at least 14-fold year over year. Across the first half of 2026 the company recorded roughly $16.2 billion in booked revenue, and the second quarter came in above the $10.9 billion figure Anthropic had earlier guided investors to expect for the period.

The most striking claim is on profitability. Anthropic said it reached positive adjusted operating income in the quarter, which would make it one of the first frontier AI labs to point to an operating profit rather than the deep losses that have defined the sector. The company cautioned that the numbers are preliminary, unaudited and could still be revised, and it declined to comment publicly on the details when contacted by reporters.

The economics behind the numbers

Underneath the headline revenue sits a business that has shifted decisively toward corporate customers. Roughly 80 percent of Anthropic's revenue now comes from its application programming interface and enterprise business rather than consumer subscriptions, according to figures shared with investors. Materials circulated during fundraising put the gross margins on that API business above 80 percent, a level far healthier than critics of the AI industry often assume.

A large part of that momentum has come from software developers. Claude Code, the company's coding assistant, passed an $8 billion annualized run rate in May 2026, reflecting how quickly professional engineers have folded the tools into their daily work. The surge in coding and agentic use cases has become one of the clearest examples of AI moving from experimentation into paid, recurring corporate spending.

Those product lines have pushed the company's overall annualized revenue, or run rate, past $47 billion as of May, according to reporting on the figures. That metric annualizes recent revenue rather than counting a booked quarter, so it runs ahead of the audited totals, but it captures the speed at which enterprise demand for Anthropic's models has compounded through the first half of the year.

The race with OpenAI

The results land in the middle of an intensifying contest with OpenAI, the rival Anthropic has been chasing for corporate accounts. OpenAI has pointed to an annualized run rate of more than $40 billion, though analysts caution that the two companies do not necessarily calculate these figures in the same way, and that run rates and booked quarterly revenue are not directly comparable measures of size.

What has changed is the perception of Anthropic itself. Once considered a safety-focused laboratory unlikely to win the commercial fight, the company has seen a steady flow of professionals adopt its software to streamline tasks such as coding, research and document work. The Q2 numbers suggest that, at least on revenue growth, the gap between the two leading American labs has narrowed considerably over the past year.

Even so, some analysts have urged caution about the profitability claim. The phrase adjusted operating income can exclude significant costs, and skeptics note that the enormous expense of training new models and securing computing capacity has not disappeared. The distinction between a genuinely profitable business and one that looks profitable only after selective adjustments is likely to be scrutinized closely once audited figures arrive.

Toward a mega-IPO

The disclosures come as Anthropic meets with investors ahead of what could become one of the largest public offerings in technology history. The company has filed confidentially for a listing and is working with Morgan Stanley, Goldman Sachs and JPMorgan Chase on the process, according to people familiar with the plans, with reports pointing to a potential IPO valuation in the region of $2 trillion later in 2026.

That would represent an enormous step up from the roughly $965 billion valuation attached to the company after a funding round in May 2026. Investors briefed on the plans have projected that Anthropic could be running at a $100 billion to $120 billion annual revenue pace by December, a trajectory that, if realized, would help justify the scale of the valuation being discussed in the market.

Why it matters

For an industry that has spent years defending vast losses as the price of building advanced models, a frontier lab pointing to an operating profit is a symbolic moment. It suggests that the business models around large language models may be maturing faster than many expected, and it strengthens the argument that at least some of the leading labs can eventually stand on their own commercially rather than relying indefinitely on fresh capital.

The open questions now are whether the audited results confirm the preliminary picture, whether margins hold up as compute costs and competition from cheaper rivals increase, and how the market receives a listing of this size. What is already clear is that Anthropic has moved from promising challenger to one of the defining commercial forces in artificial intelligence, and the second quarter of 2026 is likely to be remembered as the point where that shift became undeniable.