OpenAI filed confidential IPO paperwork with the SEC this week, according to the Wall Street Journal, with a possible public listing as soon as September and a valuation that bankers are floating somewhere between the last private round’s $852 billion and a flat trillion. The filing is confidential under the SEC’s emerging-growth-company rules, which means OpenAI gets to keep the document private during initial review and only goes fully public roughly fifteen days before the roadshow. So the most interesting financial disclosure in tech history is locked in a vault for at least a few more weeks.
The reason this matters more than a normal mega-IPO is that nobody outside OpenAI’s auditors actually knows how OpenAI’s business works. Revenue numbers leak in patches. Loss numbers leak in patches. The cost per ChatGPT query gets estimated by people who do not have access to OpenAI’s compute contracts. The Microsoft revenue-share arrangement is described in press releases as “a partnership,” which is a word doing a lot of load-bearing work. The famous capped-profit structure, which Musk’s legal team has been beating like a drum, has never been laid out as a clean diagram for the public. The S-1 is going to have to lay all of that out, in numbered footnotes, signed by the CFO.
The list of questions a serious reader will turn to first writes itself. What is the actual gross margin on inference once you back out Microsoft’s compute credits? How much of the revenue is consumer ChatGPT subscriptions versus enterprise API versus the Stargate-adjacent infrastructure deals? What does the conversion path from capped-profit to a normal C-corp actually look like, and what does it do to the Microsoft stake? How much of the next two years’ projected revenue depends on contracts that have not been signed? Anyone who has been reading OpenAI tea leaves for five years gets to finally check their guesses against reality.
The market read here is also worth pausing on. A trillion-dollar IPO of a company that almost certainly loses money on most of its operating revenue is going to be priced almost entirely on a future-AGI bet. That works in private markets where SoftBank and a handful of sovereign funds can absorb the risk without anyone tracking the daily price. In public markets, it has to clear pension funds, retail investors, and a CNBC chyron every weekday. The pricing tension between “this is the company that gets there first” and “this is a company that is still figuring out unit economics” is going to be the actual story of the fall. The S-1 is when we find out which side the numbers support.