Sam Altman was fired by OpenAI board in November 2023 and reinstated five days later after an employee revolt that made headlines globally. The episode was disorienting to outsiders — a board dismissing its own CEO only to reverse course almost immediately — but to people who had followed OpenAI closely, it made a particular kind of sense. The organization had been in tension with itself since its founding, and the firing crystallized a conflict that had been building for years between two incompatible visions of what OpenAI was supposed to be.
One vision was the one encoded in the founding documents: a nonprofit research organization pursuing artificial general intelligence for the benefit of all humanity, willing to forgo commercial success in service of safety. The other was the one that circumstances had produced: one of the most commercially successful technology companies in history, valued at hundreds of billions of dollars, in active competition with Google, Meta, and Microsoft for market position, and dependent on those same companies for the capital to operate.
Altman won. The board that fired him was reconstituted. OpenAI subsequently announced a for-profit restructuring that moved significant power away from the original nonprofit governance structure. The company that emerged from that process is something genuinely new — and understanding what it is matters for understanding where AI development is heading.
The Numbers That Changed Everything
OpenAI did not mean to become a consumer products company. ChatGPT was released in November 2022 as a research demo — a way to showcase the capabilities of GPT-3.5 to a technical audience. What happened instead was one of the fastest user adoption curves in the history of consumer technology. One million users in five days. One hundred million in two months. The scale of public engagement was so far beyond anything OpenAI had planned for that the organization had to rebuild its infrastructure in real time to accommodate it.
The consumer products revenue that followed — ChatGPT Plus subscriptions, API access, enterprise contracts — transformed OpenAI from a research organization that needed philanthropic capital to survive into a business generating multiple billions of dollars in annual revenue. That revenue created options and created obligations. It funded the compute required to train GPT-4 and subsequent models. It also created expectations from investors, from enterprise customers, and from employees whose compensation was tied to a business that now had a valuation to defend.
The Talent That Left
The history of OpenAI is partially told through the departures. Ilya Sutskever, one of the founders and the chief scientist who initially voted to fire Altman, eventually left to found Safe Superintelligence, a safety-focused research organization. John Schulman, who led reinforcement learning research and was one of the key figures behind RLHF — the technique that made ChatGPT useful — left for Anthropic, which was itself founded by former OpenAI researchers including Dario and Daniela Amodei. Greg Brockman, the president, took an extended leave.
The pattern is not coincidental. The researchers who left did not leave because OpenAI was failing. They left, in various ways, because it was succeeding — and the commercial success was creating pressures and making tradeoffs that some of the most technically accomplished people at the organization were not comfortable with. The departures represent the most sustained brain drain from a single AI laboratory in history, and the destinations — Anthropic, Safe Superintelligence, and academia — reveal the nature of the disagreement.
What OpenAI Is Actually Building
Strip away the narrative and look at the product portfolio, and what OpenAI is building becomes clear. ChatGPT is the consumer brand — the most recognized AI product in the world, with hundreds of millions of users across free and paid tiers, increasingly capable through the GPT store of specialized applications and tool integrations. The API is the developer platform — the infrastructure layer on which thousands of companies and millions of developers are building AI-powered products. The enterprise tier is the B2B business — large contracts with organizations deploying OpenAI capabilities at scale across their internal workflows.
Layered on top of this is the Operator program — partnerships with companies like Shopify, Instacart, and others that embed ChatGPT-powered agents directly into their products and services, with OpenAI capturing a share of the economic value generated. And underneath all of it is the model development that remains the core technical competence: GPT-5, o-series reasoning models, multimodal capabilities, and whatever comes next in the research pipeline that drives the commercial products.
The structure is recognizable. It is the platform playbook that Amazon ran with AWS, that Apple ran with iOS, that Google ran with Search. Build the infrastructure that others depend on, capture economics at every layer, and use the scale of adoption to generate the data and revenue that fund the next generation of capability. OpenAI is executing that playbook in AI, and it is working.
The Safety Question That Will Not Go Away
OpenAI original founding premise was that safety required keeping frontier AI research in a nonprofit that would not be subject to competitive commercial pressures. The current structure inverts that premise: frontier AI research is being funded by and increasingly directed by commercial imperatives, within an organization that has significant profit-making obligations to investors including Microsoft, which has committed over thirteen billion dollars and holds a substantial equity position.
OpenAI maintains that it takes safety seriously — it has a safety team, publishes safety research, and has implemented various guardrails in its models. Critics, including many former employees, argue that safety considerations systematically lose when they conflict with shipping schedules or commercial priorities, and that the governance structure that was supposed to prevent this from happening has been effectively dismantled.
This dispute is not resolvable from the outside with confidence. What is clear is that OpenAI is no longer the kind of organization that can credibly claim commercial pressures will not influence its development decisions. It is a business, operating with business incentives, in a competitive market. Whether that is a problem — and for whom — depends on what you believe AI safety requires and how much trust you are willing to extend to the people running the organization.
Where This Goes
OpenAI most public ambition is AGI — artificial general intelligence, a system that can perform any cognitive task a human can perform. The company believes it is on a path toward that goal, though what AGI means precisely and when it might arrive are questions where even insiders disagree significantly. What is not in dispute is that OpenAI is deploying more capable AI systems faster than any previous technology transition has moved, with less regulatory oversight than most comparably consequential industries operate under.
The empire that has been built — the users, the developers, the enterprise customers, the infrastructure dependencies — creates a form of lock-in that makes the trajectory difficult to alter even if the people running the organization wanted to. Understanding OpenAI is not just understanding one company. It is understanding the institution that more than any other is determining the pace, direction, and character of the most consequential technology transition of our lifetimes.