OpenAI generated roughly $5.7 billion in revenue during the first quarter of 2026, about $1 billion more than Anthropic’s $4.8 billion for the same period. Sherwood News reported the figure on May 21, citing earlier reporting from The Information.
The comparison deserves precision. Anthropic, per a Wall Street Journal report, is projecting that its Q2 revenue will reach $10.9 billion, which would represent more than a doubling in a single quarter. OpenAI has not disclosed its own Q2 projection publicly. The Q1 snapshot that shows OpenAI ahead does not necessarily hold past June, and neither number has been audited. Both figures are reported from sources with indirect knowledge of internal financials. Neither lab has filed public disclosures, disclosed gross margins, or disclosed how much each dollar of revenue costs to produce in compute.
What the numbers do confirm is a shared structural reality: both companies are scaling revenue faster than their compute capacity. Anthropic’s rapid enterprise adoption has been the clearest illustration of this. The breadth of its customer relationships has exposed how thin its available inference capacity actually is. Anthropic has responded by signing compute agreements with CoreWeave, Amazon, Google, and even SpaceX, spreading its capacity bets across multiple providers rather than concentrating risk in a single infrastructure partner.
The inference chip market is responding to that kind of demand. Amazon and Google have both seen strong uptake for their custom inference chips. Microsoft, by contrast, is still working to bring its Maia chips to market and has not yet landed Anthropic as an Azure customer. A Microsoft win there would validate Maia commercially and open a third scaled option for inference capacity, at a moment when every additional option has pricing power because demand is outrunning supply.
For companies building at the frontier, compute availability, not model quality and not price, is currently the binding constraint on growth. That is a meaningful shift. Eighteen months ago, the dominant question was whether demand for AI inference would materialize at scale. The question now is whether supply can keep pace with demand that has already materialized.
The figures also matter for anyone modeling the OpenAI IPO, which the Wall Street Journal reported could see a filing within weeks. Revenue of $5.7 billion in a single quarter, on a trajectory heading into a Q2 that the industry expects to be much larger, is the kind of number that anchors an IPO prospectus. The problem for would-be public investors is that none of the critical unit economics have been disclosed: not inference cost per dollar of revenue, not gross margin by product line, not the degree to which revenue is concentrated in a handful of large enterprise accounts.
Valuation context adds another layer. Anthropic’s current fundraising round carries a valuation well above OpenAI’s last reported private valuation. A company with lower Q1 revenue is currently valued higher than the company with higher Q1 revenue, which suggests investors are pricing future compute capacity and enterprise penetration, not the current quarterly snapshot.
Enterprise buyers evaluating multi-year AI infrastructure commitments should treat either lab’s revenue figure as a signal of demand validation, not cost stability. When supply is the constraint, prices do not fall on schedule. They follow capacity, and capacity is what both OpenAI and Anthropic are urgently negotiating.
Reported by Sherwood News on 2026-05-21.
