OpenAI confidentially files S-1 with SEC valuing up to $1 trillion for Q4 IPO

OpenAI has confidentially filed its S-1 registration with the Securities and Exchange Commission on Friday, May 22, with Goldman Sachs and Morgan Stanley coordinating the global operation. The target valuation is set between $852 billion and $1 trillion, according to Fortune, positioning the IPO among the three largest in American history. The public debut is expected in the fourth quarter of 2026, with September cited as the most likely date by sources close to the negotiations.

The filing coincides with the disclosure of results for the first quarter of 2026. OpenAI generated $5.7 billion in revenue between January and March, approximately $1 billion ahead of Anthropic during the same period, according to a report by The Information published this week. The growth was driven by ChatGPT subscriptions, APIs for developers, and the coding agent Codex, which accelerated the adoption of coding AI in enterprise environments over the last two quarters.

Record revenue, structural deficit

The S-1 arrives in the market with OpenAI losing $1.22 for every $1 of revenue generated in the first quarter. For the fiscal year 2025, the company reported $13.1 billion in revenue against expenditures of approximately $22 billion, resulting in a net loss of around $9 billion. Internal projections indicate an operational loss of $14 billion in 2026, driven by computing contracts with Microsoft Azure and Oracle Cloud.

The disparity between the revenue growth rate and the scale of the deficit presents investors with an unusual calculation: to evaluate a company that doubles its revenue every quarter but maintains a cost structure that will only stabilise when the cost of inference per token drops substantially. The legal quiet period prohibits public statements from Sam Altman regarding the filing until the roadshow begins.

The parallel race to the stock market

OpenAI is not the only venture capital bet in the sector preparing for a public window. SpaceX submitted its own S-1 to the SEC two days prior, on May 20, reporting revenue of $18.7 billion in 2025, a 33% increase from the previous year. At the same time, Anthropic released projections of $10.9 billion in revenue for the second quarter of 2026, with an estimated operating profit of $559 million, potentially making it profitable before finalising its own IPO.

The convergence of three substantial capital filings in less than six months, all in areas sensitive to U.S. national security, adds a layer of CFIUS review that Goldman and Morgan Stanley will need to manage before the roadshow. The agency has 45 days to review transactions with national security implications after voluntary notification, and OpenAI's role as critical infrastructure for defence and intelligence makes the review almost certain.

What the quiet period reveals to corporate buyers

The confidential filing keeps detailed financial data out of the public domain until approximately 15 days before the roadshow begins. For CIOs at consultancies reliant on OpenAI's APIs as product infrastructure, the S-1 has immediate contractual implications: public companies are required to report revenue concentration by client and may be compelled to disclose terms of agreements that currently remain confidential.

Any post-IPO price adjustments, necessary to present positive margins to public shareholders, will directly pressure the enterprise clients sustaining OpenAI’s current revenue. Long-term fixed-price contracts per token, established before the S-1 becomes public, represent the most concrete protection available for partners who have already defined their AI architecture around the company's models.

Implications for the consulting market

The IPO transforms the nature of the commercial relationship with OpenAI. Currently, the company absorbs negative margins on strategic contracts without responding to external shareholders. After the IPO, a CFO of a public company will not have that flexibility, and quarterly margin pressures will first hit clients with the highest API consumption.

AI practices that have not built parallel supply alternatives, be it Anthropic, Google Vertex AI, or open-source models such as Llama, are exposed to re-pricing risks that the capital market will make inevitable in the coming quarters. The opening of OpenAI's S-1 is not just a financial event: it is a signal that the era of market penetration pricing in enterprise AI is coming to an end.