When OpenAI told investors in February 2026 that it was targeting roughly $600 billion in total compute spending by 2030, the math was clean: four years, a defined endpoint, a number that fit on a slide. What OpenAI has not disclosed publicly is that Sam Altman, according to The Information, committed the company internally to spending $600 billion over five years — a timeline that runs to 2031 and that nobody has reported as a mismatch against the investor-facing figure.
That discrepancy is the part of the CFO-CEO rift at OpenAI that deserves more attention than it has gotten. The story is not simply that Sarah Friar, OpenAI's chief financial officer, told colleagues the company will not be ready for an IPO this year. It is that her boss has privately set a different clock than the one his company shows to investors.
OpenAI declined to comment for this article.
The reporting line shift shows part of the problem. Friar's reporting line inside OpenAI has moved: she now reports to Fidji Simo, the chief operating officer, rather than directly to Altman. That is a structural demotion. The CFO no longer has a direct line to the CEO, and in an IPO readiness process, that is not a minor operational detail — it is the person who would be building the prospectus, the financial model, and the investor narrative. Business Today first reported the shift.
The Information separately reported that Altman has been excluding Friar from certain conversations about the company's financial plans, including a key meeting involving a major investor tied to server procurement. India Today reported she was not present for a discussion about the infrastructure commitments her own financial projections depend on. People familiar with her position say Friar has pushed back on some characterizations of the rift, but the structural facts — the reporting line change, the exclusion from key investor meetings — are not in dispute.
OpenAI has told investors it is targeting roughly $600 billion in compute spending by 2030, according to CNBC. In February 2026, that was presented as a reset — a downward revision from earlier talk of $1.4 trillion for 30 gigawatts of computing infrastructure, Altman's public statement from the prior year. The $600 billion figure became the new anchor, framed as achievable within a defined period.
But The Information reported Altman committed internally to $600 billion over five years. The difference between "by 2030" and "over five years from whenever that commitment started" is not semantic. It changes the implied annual burn rate, the milestones a prospectus would need to show, and the story the company tells about when it reaches something resembling self-sufficiency. OpenAI has not issued a correction or clarification to either statement.
OpenAI expects to burn more than $200 billion before it starts generating cash, The Information reported. The company has $122 billion in committed capital and is currently generating approximately $2 billion in monthly revenue, according to LetsDataScience, which cited the same reporting. At that monthly revenue rate, OpenAI faces a burn horizon measured in hundreds of billions before anything resembling self-sufficiency.
Friar has told colleagues she is not certain whether OpenAI will need to continue pouring money into AI servers in the coming years or whether its slowing revenue growth will be sufficient to support those commitments, The Information reported. That is a reasonable question for a company at this scale. It is a difficult question to answer while simultaneously preparing for a Q4 2026 IPO.
Altman has privately said he wants to go public as soon as the fourth quarter of 2026, The Information reported. Friar does not believe the company will be ready, according to people familiar with her position. Those two facts are not a scheduling disagreement. An IPO in late 2026, against a cash burn horizon measured in the hundreds of billions, against a CFO who says she cannot yet model whether revenue will cover the compute commitments — that is the gap.
The structural problem runs underneath both issues. An IPO readiness process requires the CFO to build the financial narrative. The person being asked to build that narrative has been moved out of direct reporting to the CEO and excluded from the key financial conversations. A CFO can disagree with a timeline while remaining inside the tent. She cannot build an IPO story from outside it.
What OpenAI has not done — and what nobody has yet pinned down — is explain publicly why the investor-facing number and the internal commitment use different calendars. The company declined to comment for this article. For anyone holding OpenAI debt or equity, or trying to model the AI infrastructure buildout that OpenAI's spending represents, the difference between four years and five is not a rounding error. It is a question that has not been answered.
