OpenAI Files Confidential S-1 With the SEC, Targeting a September IPO at Up to $1 Trillion

OpenAI confidentially filed its S-1 prospectus with the U.S. Securities and Exchange Commission on Friday, May 22, 2026, formally starting the clock on what would be the largest technology public offering in history. The filing, first reported by Fortune and corroborated by CNBC, Reuters, and Axios, targets a public listing as soon as September with a valuation range between $852 billion and $1 trillion. Goldman Sachs and Morgan Stanley are leading the deal.

Because the submission is confidential, the document stays sealed until roughly 15 days before the IPO roadshow, but the financial picture leaking out is mixed. OpenAI generated nearly $6 billion in revenue in Q1 alone and is on a roughly $25 billion annualized run rate, with more than 230 million weekly ChatGPT users. At the same time, the company reported a negative 122% non-GAAP operating margin for the quarter — a loss of $1.22 for every dollar earned — translating into roughly $6.95 billion in quarterly losses on training, inference, and talent.

The S-1 will finally answer questions that years of private rounds have papered over: cost per token served, the revenue split between ChatGPT subscriptions, enterprise, the API, and Codex, and how much of the spend goes to compute versus people. Microsoft, which holds a 26.79% fully diluted stake after last year's restructuring, would see its position valued at roughly $228 billion at the low end and $270 billion at the top. The OpenAI Foundation retains about 26%, with the remainder split between current and former employees and investors.

The filing slots OpenAI into a deliberately staggered AI IPO sequence. SpaceX is set to go public in June at a $1.75 trillion valuation under ticker SPCX, OpenAI follows in September, and Anthropic — which just closed its own $30 billion round at a $900 billion valuation — is targeting October. Each listing sets a reference point for the next, but the order also exposes OpenAI to a fresh risk: the SpaceX prospectus already revealed that Anthropic alone is paying $1.25 billion a month for compute through 2029, a number public investors will now scrutinize when OpenAI's own infrastructure bill lands in the S-1.

The deeper question hanging over the roadshow is whether unit economics close before the cash does. The SEC review typically takes several months of back-and-forth on risk factors, and OpenAI's disclosures will be the first to spell out — under penalty of securities law — how the company plans to climb out of a 122% negative margin while still competing with state-subsidized Chinese labs and a Google distribution machine that doesn't need to charge per token. If September holds, Sam Altman's pitch to public markets will be the most consequential test yet of whether frontier AI is a business or a bet.