OpenAI leans into enterprise and attacks rivals

OpenAI is telling investors that the future of its business looks less like a chatbot subscription and more like software that runs inside big companies all day. On April 8, OpenAI said enterprise customers already generate more than 40% of its revenue and could reach parity with consumer revenue by the end of 2026. (openai.com) That is a sharp change from the ChatGPT story most people know, where one person opens one window and asks one question. OpenAI is now pitching “agentic workflows,” meaning chains of software agents that can pull company data, use internal tools, and complete multi-step tasks with permissions and controls. (openai.com) The company’s own language shows what it wants to sell next. Denise Dresser, OpenAI’s chief revenue officer, said companies are tired of isolated “point solutions” and want one AI layer that connects internal systems, outside data, and employee permissions in a single setup. (openai.com) OpenAI backed that pitch with scale numbers aimed at chief information officers, not casual users. The company said Codex has reached 3 million weekly active users, its application programming interfaces process more than 15 billion tokens per minute, and customers now include Goldman Sachs, State Farm, DoorDash, and Thermo Fisher. (openai.com) At the same time, OpenAI is fighting a two-front war: win corporate budgets and calm investors who are watching Anthropic gain ground. CNBC reported on April 9 that OpenAI circulated a memo to shareholders calling Anthropic “compute constrained” and “operating on a meaningfully smaller curve.” (cnbc.com) That memo was not just trash talk. OpenAI told investors it plans to have 30 gigawatts of computing power by 2030, while it estimated Anthropic would have about 7 to 8 gigawatts by the end of 2027, and Bloomberg reported that OpenAI framed its earlier infrastructure buildout as a compounding advantage. (cnbc.com) (bloomberg.com) The timing matters because Anthropic had just unveiled a locked-down cybersecurity push of its own. On April 7, Anthropic announced Project Glasswing, a program that gives a select group of partners access to Claude Mythos Preview for defensive security work. (anthropic.com) (cnbc.com) OpenAI appears to be answering that move with a more specialized product instead of a broad consumer launch. Decrypt reported that OpenAI is preparing an advanced cybersecurity offering with “trusted access” only, which would put its most sensitive capability behind a vetting process rather than releasing it widely. (decrypt.co) That tells you where this market is heading. The biggest artificial intelligence companies are no longer just racing to make the smartest general model; they are building controlled products for narrow, high-stakes jobs like software security, where access rules and audit trails matter as much as raw model performance. (openai.com) (anthropic.com) (decrypt.co) So the story here is not simply that OpenAI took a swipe at a rival. It is that OpenAI is trying to look, at the same time, like the biggest infrastructure builder, the safest enterprise vendor, and the company that can turn artificial intelligence from a chat box into the operating layer for a Fortune 500 workflow. (openai.com) (cnbc.com)