Anthropic's $11.5B quarter puts an actual P&L behind the frontier-model story

Anthropic told prospective investors it crossed $11.5 billion in preliminary second-quarter revenue, more than double the $4.73 billion it recorded in Q1 2026, and reported positive adjusted operating income for the first time ^[1][4][5]. The second fact is the one that changes how the sector gets underwritten: revenue growth at this scale was never really in doubt, but whether it could clear the cost of training and serving the models was ^[16].

The year-over-year comparison is the headline number and the least useful one. Q2 2025 revenue was $787 million, so the reported figure is roughly 14.6 times larger ^[2][1]. Sequentially it is about 2.4 times Q1, or 143 percent quarter-over-quarter growth ^[2], which is the metric that actually constrains a valuation model. Anthropic has raised more than $130 billion across its rounds, with Amazon and Google among the largest investors, and its May 2026 round priced the company at $965 billion post-money ^[10][9]. Cumulative capital raised is still more than eleven times a single quarter of revenue at the Q2 run level ^[7].

Be precise about which run-rate is which. A $11.5 billion quarter annualizes to roughly $46 billion ^[3], which is close to the approximately $47 billion annualized figure the company was citing by May ^[7] and well short of the roughly $74 billion that external trackers estimate for July 2026 ^[8]. That $74 billion is an extrapolation by third parties, not booked revenue, and it sits about $28 billion above what Q2 actually annualizes to ^[6]. On reported revenue, the May valuation is around 21 times annualized Q2 ^[4]; on the tracker estimate, about 13 times ^[5]. Those are very different comparables to hand an IPO book, and only one of them is grounded in a number the company has reported.

The ladder here has been steep even by private-market standards: roughly $14 billion annualized earlier in 2026, about $47 billion by May, with the tracker figure above that by July ^[6][7][8]. Anthropic attributes the growth to enterprise adoption of Claude and to agentic coding tools that write, debug and deploy code with limited human oversight ^[11]. The company was founded in 2021 by Dario and Daniela Amodei, both former OpenAI executives, on a stated safety and alignment mission ^[12]; it took about two years from the first major Claude release to reach nearly $1 billion a quarter, and roughly one more year to multiply that by more than ten ^[13]. At around $74 billion of annual revenue the company would sit near Intel or IBM in scale, per the source ^[15].

Three caveats are load-bearing. The revenue figure is preliminary ^[1]. "Adjusted" operating income is not defined in the disclosure, so what is being excluded, including stock compensation and any treatment of compute commitments, is unknown. And these numbers were shared with prospective investors rather than filed ^[3], which means they carry the presentation choices of a fundraising deck.

What to watch: whether an S-1 reconciles adjusted operating income to a statutory figure and discloses the compute obligations behind it; whether sequential doubling survives the next print or decays toward the annualized $46 billion base ^[3]; and whether the third-party $74 billion estimate ^[8] is validated by reported revenue. Anthropic's 14x jump implies it may be taking enterprise share from OpenAI, Google DeepMind and Meta's AI division ^[14], but share claims need two quarters of comparable disclosure from all of them, which does not yet exist.