OpenAI finance tensions

The fight inside OpenAI is not really about an IPO date. It is about whether the company can keep acting like a wartime project while also pretending it is close to becoming a normal public company. The trigger, according to The Information, was CEO Sam Altman’s push for a stock market debut as early as the fourth quarter of 2026, even as CFO Sarah Friar warned that OpenAI still lacks the financial controls, compliance systems, and operating discipline a public listing would require (theinformation.com, businesstoday.in). That disagreement would be ordinary at most companies. At OpenAI, it lands in the middle of a spending plan that is not ordinary at all. The Information reported that Altman has committed the company to roughly $600 billion in compute spending through 2030, while Friar questioned whether OpenAI would actually need that much server capacity and whether slowing revenue growth could support it (theinformation.com, sherwood.news). CNBC had already reported in February that OpenAI had reset investor expectations around a roughly $600 billion infrastructure target by 2030, down from even larger numbers Altman had floated before (cnbc.com). That number matters because OpenAI is no longer just selling software. It is trying to secure the physical substrate of the AI industry: chips, power, land, and data centers. In January 2025, OpenAI announced Stargate, a separate infrastructure effort that said it intended to invest $500 billion over four years building AI capacity in the United States, with $100 billion slated for immediate deployment (openai.com). Once a company starts making commitments on that scale, finance stops being a back-office function. It becomes the argument over what the company even is. That is why an IPO is such a loaded subject here. Public markets do not just ask whether a company is growing. They ask whether the numbers are legible. Friar’s warning, as relayed by multiple reports, was that OpenAI is still burning cash at a rate that makes a near-term listing hard to defend, especially if the company keeps layering giant compute obligations on top of a business whose growth is still real but no longer infinite (economictimes.indiatimes.com, financialexpress.com). The surprise is not that a CFO said this. The surprise is that she apparently had to say it inside a company already reorganized to make capital raising easier. OpenAI spent much of 2025 doing exactly that. By October, it had restructured its for-profit arm into a public benefit corporation while keeping nonprofit control, and Microsoft’s stake and revenue share were recast as part of the new arrangement (aljazeera.com, technologyrecord.com). In other words, OpenAI already did the corporate engineering that was supposed to make giant fundraising and, eventually, a listing more plausible. The current dispute suggests the harder problem was never the org chart. It was the math. And the math has only gotten stranger. OpenAI closed a $122 billion funding round on March 31, 2026, at an $852 billion valuation, according to Bloomberg, with major backing from Amazon, Nvidia, and SoftBank (bloomberg.com). That kind of private financing can delay an IPO. It can also make one more tempting, because once a company is valued like that in private markets, an eventual public offering starts to look less like a fundraising necessity and more like a legitimacy test. The timing is even messier because the executive bench is shifting at the same moment. OpenAI disclosed leadership changes on April 3, with Brad Lightcap moving out of the COO role into special projects and Fidji Simo taking medical leave for several weeks because of a neuroimmune condition (cnbc.com, techcrunch.com). That does not prove an IPO is off. It does make the idea of a clean, orderly sprint to public markets look harder to take seriously when the company is still rearranging who runs what while debating whether it can afford the servers it has already promised to buy.