OpenAI CFO Sarah Friar told staff at a company-wide meeting on Wednesday that the company will go public by 2027, and that the date could move earlier if growth continues to accelerate. She also addressed the question weighing most on employees' minds — that a rival has already filed its own IPO paperwork.
Slides shown at the meeting put numbers behind the message: overall revenue is annualizing at 35% growth quarter-to-date, enterprise revenue at 50%, and weekly active users for its AI coding and productivity products have reached 20 million. OpenAI's second-quarter revenue came to $6.7 billion, up 18% quarter-over-quarter, pushing annualized revenue past $40 billion.
The filing itself isn't new. OpenAI had already submitted a confidential S-1 to the U.S. Securities and Exchange Commission, with a valuation of $852 billion based on March's $122 billion funding round. Friar framed the IPO at the meeting as a milestone and a financing vehicle, not an endpoint.
The Real Question: Pace, Not Date
Employee concerns centered on one thing: Anthropic is moving faster. That company has also filed confidentially, reportedly at a $965 billion valuation, with second-quarter revenue running well ahead of OpenAI's and year-over-year growth more than doubling. Friar's answer was that OpenAI won't chase anyone else's calendar — even if Anthropic goes public within weeks of unsealing its filing and lists in September, OpenAI will move at its own pace.
"We are running our own race."
Coming from a CFO, a line like that usually does two jobs at once: reassuring staff internally, and buying room externally. IPO pricing depends heavily on the last few full quarters of financials before filing — the later the bell rings, the more high-growth quarters can be packed into the prospectus. But wait too long, and the market's valuation anchor for AI companies could shift. Both directions carry risk, which is why setting a range — 2027 at the latest, sooner if things go well — is a better bet than committing to a single quarter.
Growth Rate, Not the Calendar, Sets the Price
A rough calculation: $40 billion in annualized revenue against an $852 billion valuation works out to a price-to-sales multiple of roughly 21x. That's not out of line for the software industry — high-growth SaaS companies have historically traded even higher, provided the growth holds up. The 35% figure from the meeting is an annualized, quarter-to-date number, not a year-over-year one; the observation window is short, so quarter-to-quarter swings get amplified. What underwriters will be watching is the shape of that number across several consecutive quarters.
Enterprise revenue growing faster than the overall business — 50% versus 35% — lines up with an earlier disclosure: Friar has said at investor meetings that enterprise and consumer revenue have now "crossed," after starting the year at roughly a 60/40 split favoring consumer. For a company built on subscriptions, a shift toward enterprise revenue makes revenue meaningfully more predictable — annual contracts, per-seat renewals, low churn — exactly the kind of material that commands a premium in roadshow decks.
The 20 million weekly active users for AI coding and productivity products is another signal. A year ago, that product line was mostly a showcase for model capability; now it's being reported separately to the entire company, which suggests it has been folded into the core narrative rather than treated as an accessory to the model.
The Variables Still in Play
With the timeline set for 2027, there's plenty still to get through: the cash-flow gap created by long-term compute contracts, regulatory requirements for pre-release review of frontier models, and a more basic question — how the prospectus will explain ongoing, substantial losses. A CFO isn't going to unpack any of that at an all-hands, but each item will end up in the risk-factors section.
For readers in China, the real reference value here is the timeline. If OpenAI and Anthropic both go public around 2027, secondary markets will get their first apples-to-apples financial disclosures from AI model companies: per-unit compute cost, inference gross margin, enterprise retention — numbers that today can only be estimated from rumor will become hard, quarterly-disclosed metrics. At that point, domestic model companies pricing their own listings on the Hong Kong Stock Exchange or Shanghai's STAR Market will finally have an external yardstick to measure against.
Sources: CNBC, slide data leaked from OpenAI's all-hands meeting, CocoLoop, IT Home; revenue growth and weekly active user figures are as relayed at the meeting, and the price-to-sales multiple is the editors' rough calculation.