Anthropic and OpenAI are together generating revenue at an annualized rate of roughly $100 billion, according to Epoch AI, a research organization that tracks frontier AI development. The figure comes from Epoch’s own analysis of recent run-rate disclosures from both companies, not from audited financial statements. Sustained growth at that scale, for two companies already this large, has little precedent among software firms of the past two decades.
Run rate means recent revenue, typically from the most recent month, extrapolated to a full year. By that measure, OpenAI’s own rate nearly tripled in a year, climbing from about $13 billion last August past $40 billion today. Anthropic’s trajectory moved from roughly $1 billion to $9 billion across 2025, then kept accelerating: Epoch says the company’s run rate more than tripled again in the first quarter of 2026, and Axios reported in mid-August that Anthropic’s rate had climbed past $65 billion heading into August.
The two companies also count revenue differently, a distinction Epoch flags directly. When customers buy tokens through a cloud platform, OpenAI records only its cut of that transaction. Anthropic records the full amount, which pushes its reported figure higher relative to OpenAI’s than underlying usage alone would explain. Any comparison between the two labs’ individual numbers should account for that gap before drawing conclusions about who is actually ahead.
Combined, the two labs tripled their run rate in 2024 and expanded more than fourfold in 2025, by Epoch’s tally. This year the combined figure has already climbed from about $30 billion to roughly $105 billion through July, a 3.5x jump in under eight months. Software companies typically see triple-digit growth rates fade within a year or two of reaching product-market fit. These two labs have held that pace for three years running.
OpenAI’s own history offers a caution here. Its revenue grew roughly tenfold in 2023 after the launch of ChatGPT and GPT-4, then slowed to about threefold growth in 2024 as that initial surge diffused through the market. Epoch researchers Josh You and Lynette Bye argue the current acceleration could follow a similar arc: a temporary spike tied to coding agents crossing a usability threshold around the release of Opus 4.5, rather than a permanent shift in how fast AI revenue can grow.
None of this addresses profitability. A combined run rate of $100 billion measures how much money is moving through these two companies, not how much of it they keep. Both labs continue to spend heavily on compute and training, and neither has disclosed margins alongside these revenue figures. Revenue scale and financial health are separate questions, and treating a revenue milestone as proof of a healthy business overstates what the number actually shows.
Epoch’s own framing puts the scale in perspective. If the combined run rate triples again next year, that is one lap toward what would take roughly six consecutive years of tripling to match the scale of today’s global economy, an outcome the researchers themselves call unrealistic. Operators tracking these labs should watch the next two quarterly updates for signs the growth rate is bending: a deceleration would be the clearest evidence yet that frontier AI’s revenue curve eventually behaves like every other technology’s has.
Epoch AI, via its Gradient Updates newsletter (Josh You and Lynette Bye), published August 27, 2026.
