OpenAI confirmed last week it had submitted a confidential draft S-1 to the U.S. Securities and Exchange Commission, targeting a public listing as early as September and a valuation that could exceed $1 trillion.
The company was characteristically candid about the announcement, writing in a blog post: “We recently submitted a confidential S-1. We expect it to leak so we’re just announcing it.”
The filing comes weeks after a significant legal win. A May 18 jury verdict dismissed all claims in Elon Musk’s lawsuit against the company under California’s statute of limitations, which immediately reduced the legal pressure over the IPO. OpenAI’s bankers had reportedly been waiting for that ruling before moving forward.
The Revenue Story
The number that makes the OpenAI IPO credible is revenue growth that has few precedents in software history. According to OpenAI’s own disclosures, annual recurring revenue (ARR) scaled roughly 3x year over year, from about $2 billion in 2023 to $6 billion in 2024 to over $20 billion in 2025. By March 2026, the company was generating approximately $2 billion per month, hitting an annualized $25 billion run rate.
CFO Sarah Friar confirmed in a January 2026 blog post that the company’s ARR for 2025 was above $20 billion, against estimated full-year actual revenue of approximately $13.1 billion, which reflects rapid growth concentrated in the second half of the year.
The Cost Problem
The revenue trajectory looks impressive until you look at what it costs to generate it. OpenAI lost $1.22 for every $1 of revenue in Q1 2026. While the company is growing at extraordinary speed, it is burning capital at an extraordinary rate.
Inference costs, meaning the direct expense of running the models every time a user sends a query, reached $8.4 billion in 2025 and are projected to rise to $14.1 billion in 2026, a 68% increase in costs for what may be less than 100% revenue growth.
Every large-scale software company that traded at premium valuations shared something structural – the more customers they added, the cheaper each new customer became to serve. For instance, WhatsApp added 100 million users with nearly no additional infrastructure cost. Google’s 10th advertiser costs the same to serve as the 10th million. However, OpenAI’s model does not follow that pattern. Every additional query requires real compute, and compute is not free.
The company expects losses of $14 billion in 2026 alone, and profitability is not in sight until the end of the decade. HSBC estimates a $207 billion funding gap by 2030, which means the company may need to raise additional capital even after an IPO.
There’s also the governance question, as the S-1 will also force a clearer answer on how OpenAI is actually governed, which matters as much as the financials for anyone buying shares. OpenAI’s for-profit arm was reorganized as a public benefit corporation controlled by the nonprofit OpenAI Foundation.
What the IPO Will Reveal
The Microsoft revenue-share terms will be one of the most scrutinized sections of the S-1. OpenAI and Microsoft renegotiated their deal in May 2026, capping revenue-share payments to Microsoft at $38 billion through 2030, down from an earlier projected $135 billion. Beyond this, the full economics of their relationship, including cloud spend, IP licensing, and compute costs, will be disclosed in the prospectus for the first time.
OpenAI’s filing came roughly a week after Anthropic filed confidentially at a reported $965 billion valuation, and both are following SpaceX, which began an IPO roadshow at a valuation near $1.75 trillion. Bloomberg estimates the combined AI-related IPO pipeline is now worth around $3.6 trillion, a concentration of blockbuster offerings markets have not seen since the dot-com era.
OpenAI was explicit about keeping its options open. In its June 8 statement, the company said it had “not decided on timing yet,” adding that “it may be a while because there are things we want to do that are likely easier as a private company,” while framing the filing as preserving “the option to go public sooner if that ends up being best.” The language used mostly signals a company that wants access to public capital without rushing into the disclosure obligations that come with it.
