OpenAI spent most of 2025 launching things. A video app. A science research platform. An AI browser. A coding tool. Each one was announced with confidence. Most of them did not stick. The result was a company spread across too many products with no clear story about what it was actually building.
That approach has now been officially abandoned. Greg Brockman, one of the five people who started OpenAI a decade ago, has taken formal control of all product decisions. The head of Codex, Thibault Sottiaux, is now running the core product and platform teams. Nick Turley, who built ChatGPT into a product with 900 million weekly users, has been moved to enterprise. Three executives who led the projects being cut left the company entirely last month.
The reason for the urgency is not internal. It is competitive. Anthropic, the smaller and younger AI company founded by former OpenAI researchers, has just passed OpenAI in business adoption for the first time. Ramp, which tracks real corporate spending across more than 50,000 U.S. companies, confirmed the crossover in May 2026. A year ago, Anthropic had under 8% business adoption. Today it has 34%. OpenAI, which held 32% a year ago, is still at 32%. One company quadrupled. The other flatlined.
The engine of Anthropic's rise is its coding product, Claude Code, which has captured over 40% of the enterprise coding market compared to OpenAI's 21%. That matters far beyond software companies. Coding tools are increasingly how all businesses automate back-office work, build internal software, and connect their systems without hiring large IT teams.
OpenAI's answer is Codex, which is its own coding and task agent. Codex was described internally as one of the fastest-growing products in the company's history, and the person who built it, Thibault Sottiaux, is now running the product organisation. The entire company is being rebuilt around what Codex represents: software that does not just answer questions but takes actions, runs tasks, and completes work on a user's behalf.
The product OpenAI is now building toward is a single desktop application combining ChatGPT, Codex, and a web browser called Atlas. In practice, this means one application where a user can chat, ask it to write or edit code, browse the web, and hand it longer tasks to complete autonomously. The browser piece is what gives the agent access to the real world, so it can gather information and interact with websites without the user managing it step by step. Atlas currently only runs on Apple computers, and the full merged application has no confirmed launch date.
All of this is happening under pressure from a clock OpenAI did not set. The company is targeting a stock market listing in late 2026 or 2027, with Goldman Sachs, JPMorgan, and Morgan Stanley advising. It crossed 25 billion dollars in annualised revenue earlier this year, but is projected to lose 14 billion dollars in 2026 alone. The company does not expect to break even until around 2030. Public market investors will need a clean story about what the product is and who it is for, and until this reorganisation, OpenAI could not tell that story coherently.
Brockman now has to deliver the unified product, the enterprise revenue growth, and the IPO preparation simultaneously. Anthropic is not slowing down. It went from 9 billion dollars in annualised revenue at the end of 2025 to 30 billion by April 2026, and is reportedly weighing a funding round that would value it above 900 billion dollars. The two companies are now competing for the same enterprise contracts, the same developers, and eventually the same public market investors.
For businesses currently using any of these tools, the practical signal is straightforward. OpenAI is consolidating. Products that exist today as separate subscriptions and workflows are being pulled into one platform. The pricing and access models will change. Companies that have started adopting Codex, ChatGPT Enterprise, or the API for internal workflows should expect those surfaces to look different within the next twelve months.