Altman puts OpenAI's IPO past 2026 on a risk he says he cannot estimate

A company that needs public equity soon does not get to choose its listing year. Altman says OpenAI is not in that position: "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 told Fortune ^[3]. Reuters did not report OpenAI's cash position, so the second half of that sentence is the part an underwriter has to take on trust.

Altman ruled out 2026 and left 2027 open. Asked by Fortune whether 2026 was off the table in favor of 2027, Altman said, "I would say not 2026," and went on: "We got a lot of stuff to do, like meeting this moment of what is going to be required for safety and alignment, and how the industry and governments can work together" ^[7].

Six, eight and 10 span four percentage points. He rejected all three ^[14]. "Whether it's 10 or eight or six, the point is, we all have a tremendous amount of responsibility, and cannot let egos or incentives for profit or anything else get in the way," Altman said ^[5]. The decision he described holds at any number in that band. "We need to act such that we are not taking any of those numbers of risk, and I believe we can," he said ^[6].

The valuations in the record are both second-hand. The New York Times called the contemplated float "potentially trillion-dollar" in June ^[8], and at that time shares in Elon Musk's SpaceX were tumbling after a surge that had valued the company at $1.8 trillion ^[9]. One trillion is 56 percent of 1.8 trillion ^[15]. The larger figure is the one a public market was repricing that week.

One reading is that safety is the whole reason, and the date moves when the pact Altman hinted at, among OpenAI and other leading AI firms, actually appears ^[10]. Another is that the safety account is the public version of a decision that falling comparable valuations had already made, since the deferral question was live in June ^[8] while the nearest comparable was being marked down ^[9]. A third is that the delay is short and 2027 arrives with a governance milestone attached to it ^[7]. In my view the second is the best-supported, because the timing question predates the extinction framing by roughly three months ^[18].

That reading is wrong if OpenAI puts 2027 on the record together with the pact, because the listing date would then be following a safety agreement. It is also wrong if the labs sign what Anthropic's chief executive asked for. "We must slow the pace at which we improve the capabilities of AI models," Dario Amodei wrote in an essay on X ^[11].