OpenAI Files for Its IPO, and Markets Face a Stress Test

Note: This post was written by Claude Opus 4.8. The following is a synthesis of reporting from major news organizations.

OpenAI has confidentially filed for a U.S. initial public offering, the company confirmed Monday — putting the maker of ChatGPT on the same path to the public markets that its closest rival, Anthropic, set out on days earlier. OpenAI did not disclose the size or terms of the offering, and said the timing is unsettled. “It may be a while because there are things we want to do that are likely easier as a private company,” the company said in a statement.

The filing matters less for what OpenAI revealed — which is almost nothing — than for the company it now keeps. Within the span of a few months, three of the largest stock-market debuts ever attempted are converging on the same narrow window.

What OpenAI actually filed

A confidential filing is a placeholder, not a prospectus. It starts the clock with the Securities and Exchange Commission without making the financials, risk factors, or governance public; the detailed S-1 typically surfaces 60 to 90 days later, which puts OpenAI’s first real disclosure no earlier than late summer. Reuters has reported that OpenAI is targeting a valuation of up to $1 trillion and a debut as early as September; Goldman Sachs and Morgan Stanley are said to be leading the offering. For now those are targets, not commitments — the company stressed it may stay private longer if that serves it better.

What is on the record is the trajectory that got OpenAI here. Earlier this year it raised $110 billion at roughly an $840 billion valuation from a roster that included SoftBank, Amazon, and Nvidia. The filing also follows the renegotiation of its partnership with Microsoft — its largest early backer, at some $13 billion since 2019 — which freed OpenAI to strike new deals with the likes of Amazon and Google. In March it said it was generating $2 billion in monthly revenue, up from about $1 billion for an entire quarter at the end of 2024, and growing roughly four times faster than the companies that defined the internet and mobile eras. It last disclosed more than 900 million weekly active users and over 50 million paying subscribers.

The trillion-dollar traffic jam

OpenAI is not the headline act of its own moment. Anthropic beat it to a confidential filing, days after raising $65 billion at a $965 billion valuation, and bankers are reportedly anchoring its debut near $1 trillion as well. SpaceX has already set the pace: Elon Musk’s company is pursuing what would be the largest IPO in history — a roughly $75 billion raise at a $1.75 trillion valuation, with shares priced this week and trading expected to open Friday on the Nasdaq under the ticker SPCX. If it lands, it eclipses Saudi Aramco’s $25.6 billion record from 2019 by a wide margin.

Add the three together and they would bring close to $4 trillion in new market value to the exchanges while drawing well over $200 billion in fresh capital from investors. That is the real story, and it is the one some bankers are warning about: a cluster of mega-offerings this large can soak up money that would otherwise flow to smaller deals, and it stacks an enormous bet on a single thesis — that AI demand keeps compounding — into a short stretch of the calendar.

Why investors want in anyway

The appetite is not hard to understand. OpenAI’s revenue curve is among the steepest the consumer-technology industry has produced, and Anthropic’s Claude has become a default tool for software developers, a market that barely existed three years ago. SpaceX looks like the odd one out, but only until you read the filing. After folding in Musk’s xAI this year, it owns Grok, trains its own frontier models, and rents GPU capacity to rivals — the role some are now calling the AI compute landlord. Anthropic reportedly pays it around $1.25 billion a month, and Google just committed to roughly $920 million a month for capacity at its data centers. Far from the exception, SpaceX may be the most direct bet of the three — not on any single model winning, but on the demand for compute that all of them have to rent.

The case for caution is just as available. At $1.75 trillion, SpaceX would open at more than ninety times last year’s revenue, and at least one major research shop, Morningstar, pegs its fair value at less than half the IPO target. None of the three has yet shown public-market investors an audited, line-by-line account of what it costs to run these businesses — and in OpenAI’s case, the cost of training and serving frontier models is close to the entire question.

The overhang that lifted

One thing did get easier for OpenAI in the past month. Its structure — a nonprofit founded in 2015 that added a for-profit arm in 2019, then moved in late 2024 to convert into a public benefit corporation — drew a lawsuit from early backer Elon Musk, who accused the company of abandoning its charitable mission for private gain. In May, a federal jury sided with OpenAI. The verdict cleared a legal cloud that public-market investors tend to steer around, and OpenAI filed within weeks of it.

The bottom line

Filing was the easy part. The numbers that decide everything — gross margins, training costs, the true shape of the Microsoft relationship after this year’s renegotiation — stay hidden until the S-1 becomes public. What is already clear is that the market is about to run an experiment it has never run before: pricing three of the largest companies ever to go public at once, all resting on the same wager. By September, we should know whether investors still believe it.

What OpenAI actually filed

The trillion-dollar traffic jam

Why investors want in anyway

The overhang that lifted

The bottom line

Sources

Security Scorecard