Sarah Friar told her colleagues OpenAI isn't ready to go public. Sam Altman wants to file in Q4. They sat next to each other and called the report "ridiculous."
The Wall Street Journal dropped the story on April 28, 2026, and the AI economy bled in real time. SoftBank fell 10% in Tokyo. Oracle dropped more than 7% in US premarket. CoreWeave and AMD got dragged down with them. By the time the New York open hit, the entire AI infrastructure trade was redrawing itself around one question: what if the company at the center of all this can't pay its bills?
The Numbers Behind the Disagreement
OpenAI missed its target of one billion weekly active users for ChatGPT by the end of 2025. It also missed multiple monthly revenue targets earlier this year. Anthropic gained ground in coding. Google's Gemini gained ground in everything else. The model that built OpenAI's whole brand — being first, being best — is no longer reliably true.
Friar reportedly told colleagues she didn't believe the company was ready to IPO in 2026 because of the procedural work involved and, more pointedly, the risks from its own spending. Altman has personally signed off on roughly $600 billion in compute commitments through 2030. That's not a typo. That's six hundred billion dollars in promised payments to Oracle, CoreWeave, Microsoft, Google, Amazon, and Nvidia for chips and data center capacity.
Friar's job is to figure out where the money comes from. So far the math doesn't math.
Why the Market Panicked
The whole AI infrastructure economy was built on a simple assumption: OpenAI would keep growing fast enough to honor those compute contracts. Oracle's stock got priced for that. CoreWeave's $40 billion valuation got priced for that. Nvidia's last earnings call mentioned OpenAI by name eleven times. When the WSJ reported that OpenAI's own CFO doubts the revenue can support the contracts, every link in that chain wobbled.
Bloomberg's reporting added a detail that stung: OpenAI is now bleeding ground to competitors specifically in coding and enterprise — the two markets where you make actual money. ChatGPT consumer growth has plateaued. The free users still cost money. The paid users aren't growing fast enough to cover the data center bill.
Altman's response on the disagreement was, in effect, that there is no disagreement. Friar co-signed it. They're "totally aligned on buying as much compute as we can." Which is a fascinating thing to put in writing if your CFO has been telling people the company can't afford it.
My Opinion
I'll be blunt. This is the moment the AI hype cycle started cracking, and the crack came from inside the building.
For two years OpenAI has been treated like a country instead of a company. Sovereign-scale compute deals. Sovereign-scale press coverage. Sam Altman touring the Middle East like a head of state. The implicit promise was that demand would always run ahead of supply, so any amount of capex was justified. A billion weekly active users by end of 2025. ChatGPT as a utility, like electricity. The CFO publicly disagreeing with the CEO about whether they can pay their bills is what happens when the math finally arrives.
Here's what bugs me. Altman and Friar's joint denial wasn't really a denial. It said they agree on wanting compute. It didn't say they agree on whether the company can afford it, whether it should IPO this year, or whether $600 billion in commitments is sane. Those are the actual questions. They left them on the table on purpose.
If OpenAI does IPO in Q4, the prospectus has to disclose this exact spending picture to the SEC, with auditors signing off. If the revenue trajectory keeps disappointing, the IPO either gets delayed or gets priced at a number that nukes the secondary-market valuation. Either outcome is brutal for the people who bought OpenAI stock at $500 billion. Including, possibly, Sarah Friar's own option grants.
Watch the next earnings cycle. Microsoft, Alphabet, Amazon, and Meta all report this week. They'll announce a combined $700 billion in AI capex for the year. The market just learned what happens when one of those checks bounces.
Author: Yahor Kamarou (Mark) / www.humai.blog / 28 Apr 2026