OpenAI の 2026 年 4 四半期 IPO は実現しない可能性（9 分読了）

On April 27, 2026, the Wall Street Journal reported that OpenAI’s CFO has told other company leaders she is worried OpenAI may not be able to pay for its future compute contracts if revenue does not grow fast enough. The same reporting indicates that OpenAI’s board has been scrutinizing recent data center deals and questioning Altman’s continued push to lock in more compute capacity even as business slows. Earlier this month, The Information reported that the CFO does not report to the CEO and has been excluded from material financial discussions. Sam Altman wants OpenAI to do an IPO by Q4 2026. The market’s default read is that the IPO is a function of revenue and market conditions. Hit the numbers, pick a window, file. The company’s $852 billion private valuation, the $122 billion round, and revenue doubling are the variables that matter, and the rest is paperwork.

That read is wrong, and the reporting on Friar shows why.

There are really two problems here. The first is governance: the CFO allegedly does not report to the CEO and has reportedly been excluded from financial discussions involving server procurement, despite compute procurement being central to the company’s financial profile. The second is financing: OpenAI has committed to enormous infrastructure obligations while revenue growth, user growth, and competitive positioning are showing enough stress that the CFO and board are now scrutinizing whether those commitments can be funded.

Either problem in isolation is survivable. A company can survive an unusual CFO reporting line. A company can survive aggressive capex. A company can survive a cautious CFO. What creates the IPO problem is the interaction: unusual reporting line plus exclusion from material discussions plus public cash-burn concern plus board scrutiny plus a compressed IPO timeline.

OpenAI’s CFO, Sarah Friar, does not report to Sam Altman. She reports to Fidji Simo, OpenAI’s CEO of Applications. The arrangement has been in place since August 2025. This reporting line is not, by itself, disqualifying, but CFOs at most companies preparing for a public filing report directly to the CEO.

What is harder to dismiss is that Altman has been excluding Friar from key financial meetings. The Information reported on this in early April 2026. One such meeting reportedly involved server procurement. Server procurement is not a tangential operational matter at OpenAI. It is the central line item. The company has committed more than $1.4 trillion in infrastructure deals and plans to spend $600 billion over the next five years on semiconductors and data centers. Excluding the CFO from a major investor conversation about that exposure is not a scheduling quirk.

In November 2025, Friar told a Wall Street Journal editor on stage at WSJ Tech Live that OpenAI was looking for an “ecosystem” of banks, private equity firms, and a federal “backstop” or “guarantee” to help finance the company’s infrastructure. She and Altman subsequently walked back the federal-backstop claim. But the original comment cannot be unsaid. The comment revealed that the company’s own CFO was thinking about the infrastructure plan as something larger than ordinary private-market financing could comfortably absorb.

Per The Information’s reporting, Friar has internally raised the question of whether the company is ready to list before 2026, citing ongoing work on procedures, compliance, and organizational readiness. She has flagged the structural overlap between OpenAI’s capital suppliers and its operating counterparties — Amazon and NVIDIA appear on both sides of the table as both contributors to the $122 billion funding round and as suppliers of cloud capacity and chips.

The April 27 WSJ reporting escalates the picture considerably. OpenAI missed multiple monthly revenue targets earlier this year after losing ground to Anthropic in coding and enterprise. It missed its internal target of one billion weekly active users for ChatGPT by the end of 2025. It is dealing with subscriber defections. Friar and other executives have begun trying to control spending and establish more business discipline, putting them at odds with Altman. The board is now scrutinizing the data center deals.

A CFO worrying about cash burn is not itself a scandal. That is the job. In a normal pre-IPO company, the CFO is the person translating ambition into financeable commitments, disclosure language, internal controls, and board process. The problem here is not that Friar appears cautious. The problem is that the public record suggests a gap between the company’s largest strategic commitments and the authority of the executive who would have to explain them to public investors.

Normal CFO discipline happens inside the process. The abnormality here is the public record of exclusion from material discussions, plus public cash-burn concern, plus board scrutiny of the underlying commitments, plus a CEO actively pushing a compressed IPO timeline that the CFO is reportedly questioning.

OpenAI has now denied this twice. On April 6, after The Information’s initial report, the company told that publication that Friar and Altman “are fully aligned that durable access to compute is at the core of OpenAI’s strategy and a key differentiator as we scale.” On April 27, after the WSJ report, Altman and Friar issued a joint statement to Reuters: “This is ridiculous. We are totally aligned on buying as much compute as we can and working hard on it together every day.”

Read the second statement carefully. It denies that the two are pulling back on compute. The WSJ did not report that they were. The WSJ reported that the CFO is worried the company may not be able to pay for the compute it has already committed to, and that the board has begun scrutinizing the data center deals and questioning Altman’s strategy. The denial answers a question the reporting did not ask. The underlying record — Friar’s reporting line, her exclusion from material discussions, her on-stage WSJ Tech Live comments, her concerns about funding forward commitments, and now the board’s involvement — is unaddressed.

That OpenAI has felt compelled to issue two separate statements within three weeks denying internal discord is itself a tell. Companies preparing for IPOs generally want fewer public statements about their internal dynamics, not more, precisely because each statement creates a record that can be tested against subsequent disclosures.

The registration statement must be signed by the company’s principal financial officer, principal executive officer, principal accounting officer, and a majority of the board. Underwriters take on Securities Act Section 11 exposure for the offering document. Their defense is not vibes, growth, or strategic importance. Their defense is diligence.

That diligence becomes much harder when the public record already says the CFO was excluded from material financial discussions concerning the company’s largest forward obligations, has questioned the financing structure on stage at a Wall Street Journal event, and has, per the WSJ, told company leaders she is worried the company may not be able to pay for its compute contracts.

Counsel will not respond to that record by saying “she cannot sign.” Counsel will respond by saying remediation is required: restored authority and access, documented board review of the compute obligations, reconciled financing and revenue projections, updated liquidity disclosures, auditor comfort, and a record that underwriters’ counsel can diligence. The practical implication is the same. The S-1 cannot be filed until the remediation is in place, and the remediation cannot be parallelized with road-show prep.

If the reporting line is restored in a way that establishes a credible record of information access, that addresses one half of the governance problem. If Friar’s documented concerns about cash burn and the financing structure of forward commitments are resolved, either through restructured commercial terms or a more conservative spending posture, that addresses the financing problem. (The recent OpenAI Principles document, which doubles down on the premise that OpenAI should keep buying compute ahead of revenue because the future will justify the burn, points the other way.) If Friar herself stops pushing back on timing, that’s a tell that something has changed materially. If the board’s reported scrutiny of the data center deals produces a renegotiation rather than a rubber stamp, that’s another.

Q4 2026 is not impossible. But it is no longer just a market-window question. It requires a remediation sequence: restore the CFO’s access and authority, document board review of the compute obligations, reconcile the financing plan with the revenue trajectory, and create a record that counsel and underwriters can diligence. That sequence has not visibly started. It cannot be wished into existence during road-show prep.

If you enjoy this newsletter, consider sharing it with a colleague.

Share Buy the Rumor; Sell the News

I’m always happy to receive comments, questions, and pushback. If you want to connect with me directly, you can:

• follow me on Twitter,
• connect with me on LinkedIn, or
• send an email to dave [at] davefriedman dot co. (Not .com!)