Anthropic told investors last week that it expects Q2 2026 revenue of roughly $10.9 billion, more than double the $4.8 billion booked in Q1, and a $559 million operating profit. That would be the company’s first profitable quarter since founding. The number was disclosed inside materials for an ongoing funding round, so it lands somewhere between “official” and “the salesy version we tell limited partners,” but the revenue figure itself is solid enough that Bloomberg, CNBC, and several others ran it with sources confirming.
The growth trajectory underneath is the headline. Going from $4.8B to $10.9B in a single quarter is not a normal SaaS curve. It is the curve of a company whose enterprise contracts went GA at exactly the moment the second wave of enterprise AI procurement budgets cleared. Claude in the IDE. Claude in the call center. Claude as the backend for a thousand startups that did not exist last year. The Stainless acquisition, the Karpathy hire, and the IPO speculation that has been swirling for months all make more sense against this revenue line than they did a week ago.
The asterisks are where it gets fun. The “operating profit” definition Anthropic is using includes the cost of training new models, which is good, but it excludes stock-based compensation, which is the part where AI lab finances start to look weird. SBC at frontier labs is a structurally large number because every researcher is paid in equity that vests on five-figure share grants. Subtracting it out is technically allowed under non-GAAP reporting, and it is also the line item that, once you add it back, has historically converted “first profitable quarter” into “still losing serious money.” Ed Zitron’s newsletter ran a piece this week titled “Anthropic’s Profitability Swindle,” which gives you the dissenting view in one URL.
The other asterisk is the cost line scheduled to land in the back half of the year. Anthropic just signed an approximately $45 billion compute commitment with SpaceX’s Colossus data centers, on top of $40 billion in incremental Google Cloud spend and a Broadcom-AWS chip pipeline. The Q2 profit projection happens before most of that spend kicks in. The serious unit-economics question is not whether one quarter prints a small positive number. It is whether the revenue curve keeps doubling fast enough to outrun the compute curve through 2027. The S-1, if it shows up this year, will have the answer.