Altman delays OpenAI's trillion-dollar listing past 2026 on safety grounds

Altman put OpenAI's split nonprofit and for-profit governance in the middle of the delay. "We have put up with this incredibly complicated structure for a long time, and this moment that we're in now is kind of why," he told Fortune editor-in-chief Alyson Shontell ^[17]. Then he said what the structure is for: "We need to be able to make decisions that are not obviously in the interest of our business and our shareholders for the responsibility of fulfilling our mission and what that's going to require" ^[18]. He also said he is happy to delay the listing to do what safety requires, because "society needs to contend with these models at each level of capability" ^[15].

The conditions he named are not ones OpenAI sets by itself. Pressed on whether 2026 was off the table in favor of 2027, Altman said, "I would say not 2026. Yeah, 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" ^[3]. He said OpenAI has discussed pauses as it reaches new levels of capability ^[16], and suggested the leading AI companies may be close to announcing a pact to slow development ^[13]. Anthropic's Dario Amodei wrote in a blog post, "We must slow the pace at which we improve the capabilities of AI models" ^[12]. On Tuesday a researcher at Anthropic announced his resignation publicly and accused both Anthropic and OpenAI, where he had also worked, of acting irresponsibly in developing ever more capable systems ^[10].

There is a second explanation sitting in the same interview. SpaceX raised $85 billion, its valuation went to $1.8 trillion on the first move, and the shares then tumbled ^[7]. That raise is about 4.7 percent of the peak valuation ^[19]; the same proportion of the $1 trillion figure the New York Times attached to OpenAI in June would be a raise of roughly $47 billion ^[20]. Markets have stayed choppy since, with the Iran war re-escalating, oil prices spiking again and inflation forecasts moving up ^[8].

The two readings resolve on different schedules. A postponement driven by a broken comparison and unsettled markets ends when the comparison recovers and the markets settle. A postponement conditioned on how the industry and governments work together ends when other parties agree, and OpenAI cannot put a date on that. In my view the safety condition is the operative one, because a company that only wanted better pricing would not volunteer that its board may act against its shareholders. The counter-thesis is cleaner and may be right: SpaceX's aftermarket set the comparison, macro did the rest, and safety supplied the vocabulary. An S-1 filed in the first half of 2027 with no inter-lab agreement in place would settle it for the counter-thesis; a 2028 date arriving alongside an announced pact would settle it the other way.