OpenAI is getting more serious and less curious

The company is following the money more openly now

This is where the story gets blunt. OpenAI has given a very clear signal about where it sees the next phase of growth. In its own April 8 enterprise note, the company said enterprise now makes up more than 40 percent of revenue and is on track to reach parity with consumer by the end of 2026. In the same note, it said Codex had reached 3 million weekly active users and that GPT-5.4 was driving record engagement across agentic workflows. That is a very different kind of momentum than a speculative video app or a niche research workspace. Coding tools and enterprise infrastructure are easier to plug into recurring budgets, daily workflows, and procurement cycles. The problem is not that OpenAI suddenly stopped caring about ambitious ideas. It is that it now has clearer evidence about which ideas pay like infrastructure and which ones feel more like expensive prestige.

Coding has become the new centre of gravity

The strongest commercial thread running through all of this is coding. Reuters reported in February that OpenAI launched the Codex desktop app to gain ground in the AI coding race, particularly against Anthropic’s strength with developers. OpenAI said the app was built so people could manage multiple AI agents over long periods, rather than just use it as a one shot assistant. Since then, the official company messaging has leaned even harder into agents, enterprise workflows, and product surfaces where models become part of real work rather than side demonstrations. This is where things change. Once coding becomes the centre of gravity, everything else starts getting judged by whether it feeds that centre, can be absorbed into it, or distracts from it. Prism being folded into Codex is not just a product decision. It is a cultural signal about what now counts as core.

This is a company becoming narrower so it can become bigger

People often talk about focus as if it means shrinking. In tech, it often means the opposite. A company narrows in order to dominate a smaller number of categories more completely. That seems to be what OpenAI is doing. Reuters reported in March that top executives were finalising a major strategy shift to refocus the company around coding and business users, with leaders actively reviewing what to deprioritise. Seen through that lens, the exits of Weil and Peebles do not look random. They look like the human face of a broader organisational decision. OpenAI is trying to stop looking like a brilliant collection of experiments and start looking more like the central operating layer for work, software, and enterprise AI. That may sound less romantic than the earlier phase, but it is also the phase where very large companies usually decide what they actually are.

There is still a contradiction at the heart of it

What makes this moment interesting is that OpenAI is tightening focus while still trying to project breadth. On one hand, it is cutting standalone bets like Sora and folding science tools into Codex. On the other hand, its nonprofit arm said in March that it plans to spend at least $1 billion over the next year on AI related projects, including life sciences and medical research, and the company is still releasing highly specialised research offerings like GPT-Rosalind. That is not hypocrisy. It is the contradiction of a frontier company trying to be two things at once: a lab that wants to matter in science and society, and a business that increasingly rewards products with clearer enterprise monetisation. My view is that OpenAI is not really choosing between those identities. It is deciding which one gets to own the interface, the budget, and the daily workflow. Right now, that looks a lot more like Codex than Prism, and a lot more like enterprise agents than creative moonshots.

The risk is that focus becomes conformity

There is a strong case for this strategy, but there is also a real cost. Some of the most important breakthroughs in AI did not begin as obviously profitable product lines. They began as strange, expensive, difficult bets that looked peripheral until they suddenly did not. Peebles’ own farewell note, as quoted by The Verge, made that point in a more human way when he argued that a research lab needs room for ideas off the beaten path and warned against collapsing everything into only the most obviously important thing. That is a real tension. A company can become more disciplined and more valuable by cutting side projects, but it can also become more predictable. If OpenAI over-corrects, it may gain a tighter business and lose some of the exploratory energy that helped make it culturally important in the first place. The problem is that companies rarely notice that shift until long after it has happened.

What changes next

The next stage is likely to look less like a sprawling frontier lab and more like a vertically integrated AI company built around a few high leverage surfaces. Coding will stay central. Enterprise will keep rising as a share of revenue. Science will probably continue, but as a governed feature set, a trusted access model, or a specialised capability inside broader products rather than as a separate public identity. The exits of Kevin Weil and Bill Peebles matter because they make that transition visible. They show OpenAI shedding some of its extra skins and choosing a narrower commercial shape. That may be the right decision. It may even be the necessary decision. But it also confirms something important about the AI market in 2026: as the stakes get bigger, even the most ambitious labs start cutting away anything that does not look like the fastest path to durable scale.