OpenAI confidentially filed its S-1 with the SEC on May 22, 2026, with the company publicly confirming it on June 8. Goldman Sachs, Morgan Stanley, and JPMorgan are leading the deal. Sam Altman has publicly said he wants a September listing. CNBC is reporting a Q4 2026 window. FutureSearch, which has modeled OpenAI’s financials for two years, puts the median IPO date closer to late March 2027, once the company has more runway to show ahead of GPT-6.

The headline grabbed everyone. The financials are where it gets complicated.

Here is the number that keeps getting buried. OpenAI is running at roughly $25 billion in annualized revenue, or about $2 billion per month, growing at approximately 3.4 times the rate it was a year ago. Enterprise now makes up more than 40% of that revenue and is on track to reach parity with the consumer side by end of 2026.

Now the other side of that ledger. Full-year 2026 operating losses are projected at $14 billion on a non-GAAP basis. FutureSearch puts 2026 GAAP losses at $25 to $26 billion, roughly 80% above the widely cited headline figure, once non-cash charges and stock compensation are included. That is the number public investors will actually see when the audited S-1 goes live. Positive cash flow is not expected until around 2030. Cumulative losses between now and then could reach $44 billion.

The valuation math, as simply as possible: a company projecting $14 billion in operating losses this year is being priced at $852 billion. That is roughly 34 times forward revenue, and it assumes flawless execution across multiple years in a market where the competitive picture is changing every quarter.

Slight tangent, but it matters. The Elon Musk lawsuit against OpenAI was dismissed by a California jury on May 18, 2026, on statute-of-limitations grounds. That cleared the biggest legal obstacle to going public. It did not clear the structural ones. OpenAI still faces active litigation with the New York Times and other copyright plaintiffs. Legal analysts estimate total settlement exposure across active cases at $500 million to $5 billion, with the Times case carrying the largest variable. None of that is resolved. All of it goes into the risk factors section of the public S-1.

What’s interesting is what Anthropic does to the pricing conversation. Anthropic filed its own confidential S-1 on June 1, one week before OpenAI’s public confirmation, at a $965 billion valuation following a $65 billion Series H. Its annualized revenue run-rate crossed $44 billion in May 2026, up from $9 billion at the end of 2025. That is a fivefold increase in under six months.

More pointedly: Anthropic’s enterprise AI market share reached 34.4% in April 2026, surpassing OpenAI’s 32.3% for the first time. About 80% of Anthropic’s revenue comes from enterprise customers. OpenAI is at roughly 40%. The company expects $10.9 billion in Q2 2026 revenue and projects its first operating profit in that same quarter, roughly $559 million. A company targeting an October listing that arrives with a profitable quarter on the books, at a 20x revenue multiple, applies direct pressure to OpenAI’s 34x multiple. Investors will not miss that.

The governance layer is something public market investors rarely have to think through. OpenAI completed its conversion from a nonprofit to a public benefit corporation in 2025. The nonprofit foundation retained a meaningful equity stake and an ongoing oversight role. As of April 2026, all but one foundation board member also served on the public benefit corporation board, which raises real questions about the independence of that oversight. Public Citizen has already argued publicly that subordinating the nonprofit to the for-profit is structurally impermissible. No active legal challenge remains on that front, but the SEC will look at it closely, and not every institutional investor will be comfortable with the answer.

On infrastructure, the capital commitments are significant. OpenAI has signed a $38 billion deal with Amazon Web Services over seven years and a $300 billion total commitment with Oracle over five years starting in 2027. The company plans to spend $50 billion on computing infrastructure in 2026 alone. The Stargate project has reached nearly 7 gigawatts of planned capacity and more than $400 billion in committed investment over three years.

The IPO proceeds do not close the funding gap. They open a new, more public chapter of managing it.

The part people skip: this will be the first time OpenAI’s actual revenue mix, gross margins, customer concentration, and cost structure are disclosed publicly. Every private AI company that raised money at elevated valuations over the past two years gets stress-tested the moment that S-1 goes live. The ripple effects across venture capital, private credit, and public market AI multiples could be wider than most investors are currently modeling.

Who benefits if this works: Microsoft, which holds a 20% revenue share agreement capped at $38 billion total through 2030. Amazon, Nvidia, and SoftBank, which all backed the March 2026 round at $852 billion. Retail investors, who would gain first-ever direct access to OpenAI equity at scale.

Who loses if it doesn’t: Anyone who bought in at $852 billion expecting a near-term profitable business. The broader private AI market, which gets re-rated across the board the moment audited financials hit public markets and the gap between private marks and real unit economics becomes fully visible.

The most probable outcome is not a blowup. It is something quieter. A listing that prices at a premium, trades well for a quarter, and then forces a slow reckoning with the distance between the mission and the income statement. By the time that reckoning arrives, Anthropic may already have two profitable quarters on the books. That is the scenario nobody in the OpenAI bull case is running, and it is the one worth watching most closely.