OpenAI, Anthropic face token pricing crisis

OpenAI’s lead over Anthropic in first-quarter revenue has sharpened a separate debate in the AI market: whether frontier models are getting better faster than many customers can afford to use them. OpenAI generated about $5.7 billion in revenue in the first quarter of 2026, nearly $1 billion more than Anthropic, according to The Information, as cited by WinBuzzer and Investing.com. Anthropic, meanwhile, has gained ground with business users. Menlo Ventures said in its 2025 enterprise AI report that Anthropic had moved ahead of OpenAI in enterprise large-language-model spend and usage, based on a survey of about 500 U.S. enterprise decision-makers and its own market model. That combination — stronger adoption signals for Anthropic but a larger revenue base for OpenAI — helps explain why investors and analysts are focusing on token pricing, packaging and deployment work rather than model rankings alone. (theinformation.com) ### If Anthropic is winning some enterprise accounts, why is OpenAI still ahead in revenue? OpenAI’s first-quarter growth was tied to more than one product line. (menlovc.com) WinBuzzer, citing reporting on the companies’ financials, said OpenAI’s quarter was helped by Codex, enterprise sales and ad testing, rather than a single consumer demand spike. OpenAI also appears to have more ways to convert usage into paid revenue. (winbuzzer.com) WinBuzzer reported that the company’s subscriber base rose to 55 million by the end of the quarter from 47 million, while its broader reach gave it room to test bundles and paid conversion. Anthropic’s business has looked more concentrated around enterprise demand. Menlo Ventures said Anthropic had become the top player in enterprise AI markets in its 2025 mid-year update, with 32% of enterprise usage versus OpenAI’s 25%. (winbuzzer.com) ### What is the “token pricing crisis” people are talking about? Investing.com framed the issue as a gap between what frontier models can do and what many enterprises can budget to run at scale. (winbuzzer.com) In that view, the problem is not whether customers value the models, but whether repeated high-volume usage can fit inside normal software and IT budgets. That pressure shows up most clearly in inference-heavy products such as coding assistants, agents and embedded enterprise tools, where usage can compound quickly. (menlovc.com) Menlo Ventures said enterprise AI spending reached $37 billion in 2025, up from $11.5 billion in 2024, but also said more than half of that spend went to applications, suggesting buyers are prioritizing immediate productivity gains over long-term infrastructure bets. (investing.com) In practice, that means model vendors have to decide whether to cut per-token prices, push customers toward subscriptions and bundles, or sell more deployment and integration work around the model. That is an inference from the reported revenue mix at OpenAI and the enterprise spending patterns Menlo described. ### Why does deployment work matter so much in this fight? WinBuzzer reported that OpenAI built out a larger deployment effort, including a Deployment Company unit with about 150 engineers, to fit models into customer workflows. (menlovc.com) That kind of work can make revenue stickier because it ties model usage to internal systems, support and longer contracts. Enterprise buyers are also choosing vendors on more than raw benchmark performance. (winbuzzer.com) The reported split between Anthropic’s enterprise traction and OpenAI’s larger revenue base suggests breadth of offering, sales execution and packaging still matter. That is an inference supported by the revenue and adoption data. ### What should readers watch next? Anthropic’s next reported milestone is a $10.9 billion second-quarter revenue target cited by WinBuzzer, which said the figure could narrow the gap with OpenAI if met. (winbuzzer.com) The same report said Anthropic could post its first profitable quarter if that target holds, though it noted limited visibility into private-company accounting. OpenAI’s next clues are likely to come from pricing, enterprise packaging and whether products like Codex keep expanding inside business systems. (winbuzzer.com) For both companies, the next phase of competition looks likely to be measured not only by model quality, but by how much useful usage customers will pay for.