OpenAI confidentially filed paperwork for an initial public offering with the U.S. Securities and Exchange Commission on May 22, 2026, according to CNBC, putting a dollar figure on the AI race for the first time. The filing became widely known on June 8, when CNBC confirmed the move publicly — a week after rival Anthropic had quietly done the same thing.
The numbers being discussed are enormous, but almost none of them are locked in yet. That distinction matters more than the headlines suggest.
What OpenAI actually filed — and what it didn’t
A confidential S-1 is not a public prospectus. Under rules that trace back to the 2012 JOBS Act, a company can submit its draft registration statement to the SEC privately, work through the regulator’s comments away from public view, and only has to release the document publicly around 15 days before the investor roadshow begins. In practice, that means OpenAI’s actual revenue, losses, and cap table stay confidential for months — and the company can still delay or scrap the IPO entirely if conditions change.
That caveat is doing a lot of work here. Sam Altman has reportedly told bankers he wants the eventual public valuation to clear $1 trillion — a “hard floor,” according to Bloomberg’s reporting relayed by multiple outlets — even though OpenAI’s most recent confirmed private valuation is $852 billion, set during a $122 billion funding round that closed in March 2026. Other reporting on the filing itself puts the range under discussion at $730 billion to $852 billion. Those are three different numbers for three different things: a private mark, a filing-stage range, and an aspiration. Treat them accordingly.
Goldman Sachs and Morgan Stanley are running the deal
Goldman Sachs and Morgan Stanley are the confirmed lead underwriters, with some reports suggesting JPMorgan and Citi are also in talks to join the syndicate. It would be, by a wide margin, the largest technology IPO ever attempted — and one of the only ones where the company isn’t yet profitable.
OpenAI’s revenue run-rate has climbed to roughly $2 billion a month, or about $25 billion annualized as of mid-2026, up from $21.4 billion at the end of 2025. Enterprise customers now account for more than 40% of that revenue and are reportedly on pace to match consumer subscriptions by year-end. The other side of the ledger is less flattering: OpenAI is on track for around $14 billion in net losses this year, even as revenue grows. Wall Street has priced plenty of unprofitable tech companies before Uber and Snowflake both went public well before turning a profit — but rarely at anything close to this scale.
Anthropic changes the calculus
Timing this IPO used to be a solo decision. It isn’t anymore. Anthropic filed its own confidential S-1 on June 1, 2026, at a $965 billion private valuation, and multiple reports now put its annualized revenue at roughly $47 billion — a figure that, if accurate, would mean Anthropic has actually overtaken OpenAI on revenue despite being smaller by headcount and consumer reach. We covered how that shift happened in our breakdown of Anthropic’s revenue run-rate, and it’s the same dynamic playing out again here: Anthropic is reportedly targeting a Nasdaq debut as early as October 2026, which would put real public pricing on the table before OpenAI even confirms a date.
That’s a problem for Altman’s $1 trillion floor. If Anthropic prices first and the market decides its enterprise-heavy, safety-branded pitch deserves a premium, OpenAI could either be forced to accept a lower multiple than it wants, or watch its own listing get delayed further while it waits for better conditions. It’s worth remembering these two companies have found common ground before — both were signatories, alongside Google DeepMind, on a rare joint statement on AI oversight earlier this year — but on IPO timing, this is a straight competition for the same pool of institutional capital.
Why the September date is already shaky
OpenAI’s original internal target was a September 2026 listing. Reuters reported in late June that the company may now wait until 2027 instead, citing market volatility and Altman’s reluctance to price below his $1 trillion target. Nothing about that has been confirmed one way or the other since — which is normal for a confidential filing. Companies routinely use the process to test investor appetite without committing to a date, and both September 2026 and a slip into 2027 remain live scenarios.
The legal backdrop got a little cleaner right before the filing: a jury dismissed Elon Musk’s lawsuit against OpenAI just two days earlier, removing one of the more visible overhangs on the company’s path to going public. It’s not the only litigation OpenAI is dealing with — the company is still fighting a separate, escalating copyright case, one we detailed in our coverage of the New York Times lawsuit — but it’s one fewer variable for bankers to price around.
What to watch next
Three things will tell you more than any leaked valuation number: whether OpenAI’s public S-1 actually surfaces in the 15-day window ahead of a roadshow (which would confirm a real date is set), whether Anthropic prices first and at what multiple, and whether OpenAI’s enterprise revenue keeps closing the gap with consumer subscriptions fast enough to justify a trillion-dollar ask to public-market investors who, unlike venture backers, will see the burn rate every quarter.
Until the public filing lands, every number attached to this IPO — the valuation, the date, even the “confidential” underwriter list — should be read as reported, not confirmed.
Deixe um comentário