OpenAI Files Confidentially for an IPO
The confidential submission, reported in June, starts a clock on what could become the largest listing since SpaceX — and forces the company to answer questions it has avoided as a private business.
OpenAI submitted a confidential draft registration statement to securities regulators in June, according to reports, beginning the formal process toward an initial public offering that bankers have spent two years anticipating.
A confidential filing is not a commitment. It lets a company begin the review process privately, resolve accounting and disclosure questions with regulators, and decide later whether and when to proceed. Executives have pointed to a window running from late this year into 2027.
The context: a market that has proven it can absorb size
The listing environment could hardly be more favorable. U.S. issuers raised roughly $251 billion in the first half of the year, a record. In June, SpaceX raised about $75 billion in the largest offering in history and closed its first day up 19%.
That transaction removed the ceiling. Before it, the largest deal ever completed had raised about $25.6 billion; after it, bankers stopped telling clients what the market could not handle.
The company's most recent private valuation was around $852 billion, set in a funding round of roughly $122 billion that closed in late March. Its closest competitor has since raised at a valuation approaching $965 billion.
What the filing forces into the open
A registration statement requires disclosures a private company can decline to make: audited financials, the true cost of compute, contract terms with cloud and chip suppliers, customer concentration, and the accounting treatment of enormous long-term infrastructure commitments.
It also requires a governance description. The company's structure — a capped-profit arrangement operating alongside a nonprofit — has been the subject of years of debate, and a listing requires it to be explicable to public shareholders in a document that carries liability.
The question public investors will ask
Revenue has grown very fast, with enterprise now a large and rising share of the mix. Costs have grown alongside it, because training and serving frontier models consumes compute priced by a supplier with extraordinary pricing power.
The bull case is that this becomes the platform layer of a general-purpose technology. The bear case is that model capability is converging, that competitors are closing the gap, and that the business ends up as a very large infrastructure operator with the margins that implies rather than the ones a $1 trillion valuation requires.
