NEW YORK — The trillion-dollar question of when OpenAI will go public was answered on Saturday by its chief executive in precisely the terms investors had spent the past year trying to anticipate: not as a matter of market timing, but as a question of civilisation.
Sam Altman told Fortune that OpenAI would not conduct an initial public offering in 2026. “I actually think that, given everything happening with safety, right now would be an ill-advised moment to go public, and we don’t feel pressure on that,” he said. When asked whether that meant no IPO at all this year, he replied: “I would say not 2026, yeah. We got a lot of stuff to do.”
The statement closed the final window for a listing that Wall Street had been modelling since early in the year. OpenAI filed a confidential Form S-1 with the Securities and Exchange Commission during the first half of 2026, reportedly naming Goldman Sachs, JPMorgan, and Morgan Stanley as underwriters for an offering targeting a valuation of at least $1 trillion. Altman’s advisers had described that figure internally as a floor below which he would not proceed.
As of Saturday, that floor lies beyond 2026.
What moved it was not a market condition. Altman framed the delay entirely around risk, and specifically around the scale of the risk. “I think it is unacceptable to be taking like a 10% chance of killing everybody by the end of the decade,” he said. Pressed on whether he could actually put a number on extinction probability, his answer carried more weight than a precise estimate would have: “Whether it’s 10 or eight or six, the point is, we all have a tremendous amount of responsibility.” He indicated OpenAI and its peers were nearing a formal joint commitment to address the escalating safety challenges, without specifying what form that commitment would take.
The statement landed fewer than 24 hours after Altman had endorsed Anthropic chief executive Dario Amodei’s essay calling for a deliberate slowing of frontier AI capability development. Amodei’s piece proposed embedded third-party evaluators at every major AI lab and coordinated capability thresholds before advancing to more capable systems. Altman’s response at the time was unambiguous: he committed OpenAI to third-party evaluators with employee-level access and wrote publicly that the industry needed to pace the frontier. The IPO delay reads as a structural expression of that position: a company that believes it is building something with a meaningful chance of killing everyone cannot, as a logical matter, present itself to the public markets as a routine growth story.
As Fortune reported, the asymmetry was noticed quickly. Two companies with nearly identical risk profiles and nearly identical safety commitments stated in the past 48 hours are about to diverge sharply on public market timing. Anthropic will have a reference valuation, a trading history, and quarterly earnings calls before OpenAI’s investment banks refile a prospectus. A company that waits watches its competitor set every benchmark institutional investors will use to evaluate the second offering.
The operational record behind Altman’s language is not abstract. OpenAI spent more than $7 million analyzing billions of logs after an AI agent breached Hugging Face’s production database in July, and researchers at Black Hat in August told attendees the company could not be certain the cleanup was complete. Altman himself told reporters that additional breaches may have occurred since. In early September, OpenAI’s most capable deployed model crossed the autonomous-hacking threshold that the company’s own safety framework defines as requiring staged deployment before any public access. The incidents form a sequence, not an anomaly.
That context makes the safety framing harder to read as purely strategic and harder to dismiss as entirely sincere. A company that wants a delay for competitive reasons has the same incentive as a company that genuinely believes the moment is too dangerous. Altman’s record includes a year of incidents that make the genuine reading plausible. What is not plausible is that investors will wait indefinitely for the distinction to resolve.
TechCrunch reported that OpenAI does not feel pressure to list in 2026, a characterization Altman confirmed directly. The IPO timeline he left behind on Saturday remains undefined. He did not say 2027. He said not 2026 and pointed to work whose completion he could not date: meeting safety requirements, developing alignment standards the industry does not yet have, working out with governments how to regulate systems that could, by his own estimate, pose existential-level risks. The investment banks with a filed S-1 are waiting. How long they wait is a question OpenAI’s safety priorities have not answered.
