OpenAI is assembling the machinery for what would be the largest technology IPO in history. It has been working with Goldman Sachs and Morgan Stanley on a confidential filing and is targeting a public debut as soon as September, with a fourth-quarter fallback if markets turn. The valuations being floated are staggering and, tellingly, all over the map: reporting puts the range at roughly $730 billion to $850 billion, with some accounts reaching toward $1 trillion. OpenAI is said to be aiming to raise around $60 billion — a haul that would more than double the $25.6 billion Saudi Aramco pulled in 2019, still the record. Whatever the final number, a company that had no product a decade ago is about to ask the public market to price the entire AI thesis.
Here is the part that complicates the pitch. The rival OpenAI spent two years out-running has quietly passed it on the one metric an IPO is supposed to reward. Anthropic, the maker of Claude, is running at roughly a $47 billion annualized revenue rate as of the spring — up from about $1 billion at the start of 2025 — while OpenAI, by its own accounts, sits somewhere between $25 billion and $33 billion. Fortune reported the crossover in early July. The two got here by selling to opposite ends of the market: Anthropic books around 85% of its revenue from enterprises and developers wiring Claude into their software, while OpenAI earns a similar share from ChatGPT consumer subscriptions. One curve compounds quietly through corporate contracts; the other rides on 50 million-plus consumers who can cancel any month.
That split is the real question underneath the offering. OpenAI has never turned a profit — internal projections point to a roughly $14 billion operating loss this year even as revenue climbs — so the valuation is a bet on the durability of consumer demand, not on today’s cash flows. Buying in near $850 billion means believing ChatGPT’s lead in mindshare converts into pricing power before the losses compound. The competition is not making that easy: cheaper frontier models keep undercutting the leaders on price, and OpenAI walks toward the market carrying an Apple trade-secret lawsuit as a live risk factor.
Our take: The revenue crossover doesn’t make OpenAI’s IPO a bad deal — it reframes it. Public investors are being asked to hang the highest price tag in tech history on the AI company with the lower revenue and the heavier burn, precisely because its brand reaches the most people. That is a coherent bet, but it is a bet on consumer AI as a durable franchise, not on financial fundamentals. Anthropic’s enterprise engine is worth watching for the opposite reason: it is starting to look like a software business rather than a story. The tell won’t be the opening-day pop. It will be whether, a year from now, the consumer moat still commands the premium the private markets have been paying for it.
What to watch
- The price, not the debut. A September listing near the top of the range would be a vote of confidence in consumer AI. A quiet slip to the fourth quarter, or a trimmed valuation, would say the market blinked.
- Revenue mix in the filing. When the prospectus goes public, the split between consumer subscriptions and enterprise/API revenue is the number to find — it decides whether OpenAI gets valued like a software business or like a media franchise.
- The loss curve. A projected $14 billion operating loss for 2026 is the counterweight to every growth chart. Watch the path to breakeven the company lays out, and whether it is believable.
- Bubble nerves. The listing lands while central banks are openly warning about AI over-leverage. A record IPO into a jittery tape is its own kind of stress test.