OpenAI's Developer Day in San Francisco on 29 September 2026 was meant to be about software — new tools, new capabilities, new toys for developers. The conversation that travelled furthest, however, came from a sit-down with CNBC's Kate Rooney in which chief executive Sam Altman addressed the question markets have been asking all year: when will OpenAI go public? His answer was blunt in its vagueness — and revealed how the company now frames its own timeline.
Meridian did not conduct this interview. This feature is an editorial reading of Altman's published on-record remarks, with every quotation taken from the coverage cited in the sources below.
Altman spoke against a backdrop of intense speculation about a listing, months of safety debates across the industry, and a recent decision by OpenAI to hold back a model — a move that had set tongues wagging about what, exactly, the company had found. His message across the interview was consistent: safety and mission come before the stock market.
No particular timeline for an IPO
Asked directly about OpenAI's plans for an initial public offering, Altman closed the door on any near-term expectation:
"I don't have a particular timeline in mind."
— Sam Altman, in a September 2026 interview with CNBC at OpenAI's Developer Day
He went further, arguing that listing while the technology and its guardrails are still in flux would be a strategic error. An IPO, he suggested, could only become realistic once the company has run "a few safety cases" and feels it understands how to safely contend with the "next level of AI".
"Barreling forwards with an IPO, while I think things feel so in flux, and I really want to be able to just focus on the mission and not with all of the challenges of being a newly public company, I just, I don't want to do that. I really think this is a time to put safety and mission first."
— Sam Altman, as reported in CNBC's live DevDay coverage (29 Sep 2026)
For a company that would instantly rank among the most valuable listings in history, the stance is striking. Altman is effectively telling investors that OpenAI's readiness for public markets is a function of its scientific understanding of the next generation of models — not of revenue run-rates or market windows. Whether public shareholders would accept "a few safety cases" as an IPO precondition is a question the company has now deliberately put on the table.
Why OpenAI held back a model
The interview also gave Altman his fullest public explanation yet of a recent decision to withhold a new model — a move some observers had read as a sign of something alarming inside the lab. Altman worked hard to deflate that reading:
"There's not a big scary thing in this case."
— Sam Altman, in a September 2026 interview with CNBC at OpenAI's Developer Day
The model, he said, was "a little bit worse on a few of the evals" the company uses to gauge safety and alignment — a shortfall that triggered what he described as an "abundance of caution". He presented withholding or reworking models that don't meet OpenAI's standards as part of the normal development process, not an emergency brake.
The framing matters beyond this one model. By describing safety gates as routine engineering discipline, Altman is trying to normalise an industry in which delayed releases are a sign of process working — a message aimed as much at regulators and the public as at developers.
Safety as science, not just engineering
Perhaps the most consequential idea in the interview was Altman's resistance to the notion that AI safety is simply a matter of better engineering. Alignment, he argued, remains an open scientific problem:
"We still have discovery about how to align these models."
— Sam Altman, in a September 2026 interview with CNBC at OpenAI's Developer Day
More broadly, Altman said alignment, safety, monitoring and security need to remain "way ahead of capabilities" as models improve — an approach he described as "pacing" progress. On the question of who is liable when an AI system fails, he said he expects a multi-level framework to emerge, depending on whether a failure originates in the model, in how a developer uses it, or in intentional misuse by an end user — comparing the possible structure to the automotive industry, where responsibility can fall on different parties depending on what went wrong.
Taken together, this is an argument for a slower, more layered governance of AI development: scientific discovery first, engineering second, and liability frameworks built for a world where failure can be nobody's fault in particular.
Not worried about being first on hardware
The DevDay conversation also touched on AI hardware — an arena where rivals, including Meta, have been making aggressive moves. Altman teased that there could be "something that's worth waiting for" on the horizon, but dismissed the race to be first:
"I am not worried about being first with new hardware. New hardware is very hard to do. New computing form factors come along very rarely, and the quality bar that we want to hit for something that you will love and really be a key part of the way you interface with AI, that's going to take a while, and we're going to do that at a high level of quality."
— Sam Altman, as reported in CNBC's live DevDay coverage (29 Sep 2026)
It is a classic Altman formulation — patience as strategy. Computing form factors, he argued, arrive rarely, and the bar for something people will genuinely love is high. The company would rather arrive late and excellent than early and forgettable.
Washington, and the anxiety economy
Finally, Altman addressed the two meta-questions that hover over every public appearance: regulation, and the public's unease about AI itself. On Washington, he was cautiously optimistic — with a caveat born of experience:
"I have been wrong before when I thought something was going to happen in D.C., and, you know, it didn't. But I do feel like people are taking it seriously this time, and in spite of previous experience, I'm optimistic."
— Sam Altman, as reported in CNBC's live DevDay coverage (29 Sep 2026)
And on the cultural anxiety that has made him, personally, a lightning rod — including "Artificial", a forthcoming film about his dramatic 2023 ouster — he was relaxed:
"I mean, looks like an awful movie, but I appreciate people's creativity, and I really do support that. So I hope they have a fun time with it."
— Sam Altman, as reported in CNBC's live DevDay coverage (29 Sep 2026)
He added that he understands why people have anxiety about AI, and that if they need to project that anxiety onto him for a little relief, "it doesn't bother me". It was the most human moment of a highly strategic interview — and perhaps the most revealing. Altman's OpenAI is asking the world for patience: on the IPO, on the models, on the hardware, and on itself.
Sources
