Bloomberg reported Friday that OpenAI is now considering a 2027 IPO and expects rival Anthropic to take the public market first, even though both companies have already submitted confidential S-1s to the SEC. Anthropic filed on June 1 with a target October 2026 Nasdaq debut at a $965 billion valuation, with Goldman Sachs, JPMorgan, and Morgan Stanley as joint underwriters. OpenAI was targeting a fall 2026 listing as recently as last quarter. That window is now reportedly closed.
The proximate reason is the valuation. Sam Altman is reportedly refusing any listing that prints below one trillion dollars. CFO Sarah Friar is reportedly pushing internally for the 2027 timeline to give the equity story time to compound around the Jalapeño inference cost reduction and whatever GPT-5.6 generates in enterprise renewal revenue. The two positions are functionally the same position. OpenAI cannot list at $965 billion eight weeks after Anthropic prints the same number, because the headline writes itself. Either you are bigger than the smaller company or you wait until the public market lets you charge a premium that makes the size difference visible on the screen.
The deeper reason is that the public market is not the private market. Anthropic’s $965 billion target is roughly triple where the company priced eighteen months ago, which is a real number underwriters can sell to institutional buyers without anyone having to invent a new framework. OpenAI’s last private mark was reportedly $730 billion. A trillion-dollar IPO would require the public market to underwrite a 37 percent step-up over a six-month-old private valuation that itself was set during peak capital-supply conditions. SpaceX, which has been quietly used as the cautionary comp in OpenAI’s internal deliberations, is a useful reminder that companies the public market thinks it wants can still trade sideways for years when the most exciting line on the prospectus is “valued at.”
The competitive read is that this is the first time OpenAI has been forced to react to Anthropic’s pacing rather than the other way around. For three years the default model was that OpenAI set the calendar and every other lab marked to it. Anthropic filing first and dragging the IPO window into October has flipped that arrangement. OpenAI’s response, which translates as “we will go later at a bigger number,” is the response of the company that lost the first move. Whether it is the right response depends on whether the public-market comp set for AI infrastructure is bigger or smaller in eighteen months than it is today. The honest answer is that nobody knows, including the people in San Francisco refusing to file.
The thing to watch between now and October is what Anthropic’s book actually builds at. If the deal prices at a premium to $965 billion, the trillion-dollar OpenAI thesis gets meaningfully easier to underwrite and Friar’s 2027 framework looks like patience. If it prices at a discount, or breaks below issue on day one, Altman’s hold-out becomes the most expensive scheduling decision in the AI sector so far. The order of operations is now the story, and the order of operations was supposed to be the one thing OpenAI got to keep.