OpenAI projects $14B 2026 loss and signals IPO postponed to 2027

OpenAI’s problem is not demand. ChatGPT is huge, enterprise AI spending is real, and investors still treat the company like one of the defining businesses of this cycle. The problem is the bill. Internal reporting that surfaced in late April and early May says OpenAI is staring at roughly $14 billion in losses in 2026 on about $13 billion in revenue, and CFO Sarah Friar has been arguing that an IPO should wait until 2027, not 2026. ### What actually broke here? The clean growth story cracked. The recent reporting says OpenAI fell short of some of its own revenue and user-growth expectations, which matters because the company’s spending commitments were built for a much steeper ramp. If growth comes in merely excellent instead of absurd, the financing math changes fast. (cnbc.com) ### Why are the losses so big? Because AI is a weird business — software margins on the surface, industrial-capex habits underneath. OpenAI has to pay for training, inference, data centers, chips, and long-term compute contracts at a scale that looks more like infrastructure than SaaS. That is why a company can post eye-watering revenue and still burn eye-watering cash. (cnbc.com) ### Why does the IPO timing matter? Going public is not just a fundraising event. It is a credibility test. Friar’s reported argument is basically that public markets would punish a company that still has this much uncertainty around spending, runway, and the path to positive cash flow. Sam Altman, meanwhile, has reportedly wanted a faster timeline. That tension matters because it tells you OpenAI is deciding whether to sell a dream now or wait for cleaner numbers. (cnbc.com) ### Is $13 billion in revenue still impressive? Yes — extremely. In most industries, that would end the argument. But the catch is that OpenAI is being judged against a valuation story that implies much more than “fast-growing software company.” Once expectations drift toward trillion-dollar territory, investors stop asking whether revenue is big and start asking whether the business model compounds without swallowing all its own gains. (ibtimes.co.uk) ### Why is compute the whole story? Because every extra user query is not free. Generative AI can keep scaling demand and still struggle economically if serving that demand requires huge ongoing hardware and power spend. That does not mean the business is broken. It means the company has to prove that newer models, better routing, enterprise pricing, and infrastructure deals can bend costs down faster than usage pushes them up. (cnbc.com) The market is no longer giving that for free. ### Does this mean OpenAI is in trouble? Not in the simple sense. A company can lose a lot of money while building a dominant platform. Amazon trained investors to tolerate that. But Amazon’s story worked because the underlying economics got better with scale. OpenAI now has to show the same thing in an AI context — that today’s spending spike is laying down a moat, not just financing a very expensive race. This is the real question behind the IPO debate. (cnbc.com) ### What changes if the IPO slips to 2027? Mostly, the burden shifts back to private investors, strategic partners, and debt or structured financing. That buys time. But it also means OpenAI has another year in which every new model launch, enterprise contract, and infrastructure deal gets read through one lens: is the business becoming more efficient, or just bigger? (ibtimes.co.uk) ### Bottom line? The headline is not that OpenAI suddenly stopped growing. It is that growth no longer settles the argument. The company still looks like one of the strongest AI businesses in the world. But the latest numbers make the core issue impossible to ignore — in generative AI, scale is easy to imagine, and much harder to afford. (ibtimes.co.uk)