OpenAI delays planned IPO to 2027, CFO cites funding and product gaps

OpenAI’s IPO story is getting framed like a simple delay. But that’s not quite the real thing. The actual news is that Sarah Friar, OpenAI’s CFO, is reportedly arguing that a late-2026 public listing is too aggressive, while Sam Altman has been more open to moving faster. The reason is not one bad quarter. It’s that OpenAI is trying to finance a company that looks part software business, part infrastructure builder, and maybe part hardware lab at the same time. (gizmodo.com) ### Is there actually a confirmed 2027 IPO? Not officially. OpenAI has not published an IPO date, and it has not publicly said, “we are delaying to 2027.” What exists is a cluster of recent reports saying Friar favors waiting until 2027 because the company is not yet where a public-market CFO would want it to be — espe(gizmodo.com)cleanly to investors. (studioglobal.ai) ### Why would Friar want to wait? Because public markets punish messier stories than private markets do. A private investor can buy the “winner takes most” AI argument and tolerate huge losses for a while. A public investor still wants that story, but also wants cleaner segment reporting, more pr(studioglobal.ai)ake future compute contracts harder to support. (msn.com) ### What’s the spending problem? The number that makes this feel real is $50 billion. Greg Brockman said OpenAI expects to spend that much on computing in 2026 alone. Layer that onto earlier reporting that internal forecasts pointed to roughly $14 billion in losses in 2026, and you get the basic tension: OpenAI is still growing fast, but the machine underneath that growth is absurdly expensive to run. (bloomberg.com) ### Why is compute so expensive here? Because OpenAI is no longer just renting cloud capacity like a normal software company. It is tied to a much bigger infrastructure push. In January 2025, OpenAI announced Stargate, a project that said it intended to invest $500 billion over four years building A(bloomberg.com)the company’s future is now entangled with data centers, power, chips, and long-term capacity deals. (openai.com) ### But didn’t OpenAI just raise a ton of money? Yes — and that’s why this is not a liquidity panic. OpenAI said in March that new funding and partnerships materially strengthened its balance sheet, and CNBC reported the latest round closed at an $852 billion post-money valuation. SoftBank separately said it agreed to invest another $30 billion, bringing i(openai.com)an still raise private money. The question is whether it wants public investors looking at the hood right now. (openai.com) ### Is growth actually slowing? It depends on which layer you mean. Consumer reach is still enormous — TechCrunch reported OpenAI said ChatGPT had reached 900 million weekly active users in February. But recent reports also say OpenAI missed some internal revenue and user-growth targets. That’s the awkward combination Friar seems worried about: huge scale, huge hy(openai.com)structure commitments of this size. (techcrunch.com) ### Why does hardware make an IPO harder? Because hardware and infrastructure bets muddy the software multiple. Investors love recurring software revenue. They get more cautious when the same company also needs to explain custom chips, robotics ambitions, data-center leases, power contracts, and uneven payback periods. The catch is th(techcrunch.com)e a capital-intensive industrial project, the harder a clean IPO pitch becomes. That last part is an inference from the spending mix and reported internal debate. (bloomberg.com) ### So what’s the bottom line? This looks less like a classic delay and more like a fight over what OpenAI is becoming. If it were just a software company, late 2026 might be plausible. But it’s trying to be the model maker, the product company, and the compute buyer of last resort all at once. Friar seems to want one more year to make that story legible. (studioglobal.ai)