OpenAI: Enterprise Is 40%

OpenAI says business customers now bring in more than 40% of its revenue, and it says that share could match consumer revenue by the end of 2026. Two years ago, the company was still known mainly for a chatbot used by individuals. (openai.com) That change means large companies are no longer treating artificial intelligence like a demo on a side screen. OpenAI’s own enterprise report says more than 1 million business customers already use its tools, which is what “pilot project” starts to look like when it turns into regular software spending. (openai.com) The shift inside companies is from asking one chatbot one question at a time to giving software a job and letting it work through steps on its own. OpenAI describes this as “agentic workflows,” where tools can plan, write, check, and hand work back, more like a junior employee than a search box. (openai.com) The clearest place to see that is coding. OpenAI says Codex, its software engineering agent, has reached 3 million weekly active users, and the product is built to handle features, refactors, reviews, and releases across multiple parallel work threads. (openai.com) That number moved fast. Three weeks earlier, OpenAI said Codex had over 2 million weekly active users, which means the tool added at least 1 million weekly users in less than a month. (openai.com, openai.com) OpenAI has been building the product around that kind of use, not around one-off code suggestions. When it introduced the Codex app in February 2026, it called it a command center for multiple agents running long tasks in parallel across projects. (openai.com) The company is also using customer names to show this is not just start-ups buying seats. OpenAI says new customers include Goldman Sachs, Philips, and State Farm, while existing users include DoorDash, Thermo Fisher, Cursor, and LY Corporation. (openai.com) Under the hood, OpenAI says its application programming interfaces now process more than 15 billion tokens per minute. That is the plumbing number behind the headline: companies only run agents in production when the system is fast enough and cheap enough to sit inside real workflows all day. (openai.com) OpenAI’s own case studies are starting to read less like experiments and more like operating manuals. In one March 2026 example, the company said a small three-engineer team using Codex opened and merged roughly 1,500 pull requests into a codebase of about 1 million lines over five months. (openai.com) The business backdrop matters here too. In June 2025, OpenAI said it had 3 million paying business users; by January 2026, it said it had more than 1 million business customers overall; and by April 2026 it was saying enterprise had become a revenue engine large enough to rival consumer. (venturebeat.com, openai.com, openai.com) So the story is not just that OpenAI sold more corporate contracts. It is that the part of artificial intelligence that writes code, routes tasks, and plugs into company systems is now big enough that OpenAI says it accounts for roughly two dollars out of every five the company makes. (openai.com, decrypt.co)