Leaked OpenAI internal figures show roughly $25B revenue but a projected ~$14B operating loss

OpenAI is no longer just an AI lab story. It is a capital structure story. The company appears to be generating revenue at a scale software investors usually dream about — roughly $25 billion annualized by early 2026 — but the catch is that the spending curve is still so steep that even that kind of top line may not cover the bill. Then, in late April, reports that OpenAI missed internal sales and user targets brought the underlying tension into view. (bloomberg.com) ### Why are people suddenly focused on the numbers? Because the growth narrative cracked a little. The late-April reporting said OpenAI fell short of internal goals for both revenue and new users, and that finance chief Sarah Friar had raised concerns about whether the company could comfortably support its future compute commitments if growth stay(bloomberg.com)s were already primed to worry about AI economics, not just AI demand. (money.usnews.com) ### What does the $25 billion figure actually mean? It does not mean OpenAI booked $25 billion in profit, or even necessarily full-year realized revenue. The figure circulating in reporting is annualized revenue — basically a run rate extrapolated from current business momentum. Bloomberg reported in February th(money.usnews.com)roducts, APIs, and newer businesses. (bloomberg.com) ### Then why is a $14 billion loss such a big deal? Because it shows how brutal AI cost structure is at the frontier. A normal software company can scale revenue much faster than costs once the product works. OpenAI is different. Every jump in model capability drags along more chips, more data centers, more inference expense, and more long-term in(bloomberg.com)d to internal forecasts has put that 2026 operating loss around $14 billion. (finance.yahoo.com) ### Is this just an OpenAI problem? Not really — but OpenAI is the clearest example. The whole frontier-model business is discovering that “software margins” do not automatically apply when the product is built on scarce GPUs and giant cloud commitments. OpenAI just happens to be the company doing it at the biggest scale, with the most visibility, and with investors that have already committed extraordinary sums. (bloomberg.com) ### How much money has it already raised? A lot — and that is part of why this matters. OpenAI finalized a $40 billion round in March 2025 at a $300 billion valuation, then completed an even larger $122 billion round in March 2026 that valued the company at about $852 billion. Those financings bought time and compute. But they also raised the bar. Once private investors are writing checks that large, they expect a believable path from growth to durable margins. (bloomberg.com) ### Why does missed growth make the burn scarier? Because the whole model depends on outrunning fixed commitments. If revenue is compounding faster than compute obligations, the losses can be framed as temporary. If growth slips while infrastructure contracts stay locked in, the economics start to look less like Amazon-in-1999 and more like a company financing an arms race. That is why even a modest miss versus internal targets landed so hard. (money.usnews.com) ### So what are investors really debating now? Not whether OpenAI has product-market fit. That part looks real. The debate is whether the company should keep spending as if scale itself is the moat, or tighten compute discipline before public-market scrutiny forces it to. IPO timing, private-credit bridges, and (money.usnews.com) effectively asking investors to fund several more years of imbalance. (bloomberg.com) ### Bottom line? The important leak is not “OpenAI is losing money.” Everyone assumed that. The important leak is that even at roughly $25 billion in revenue, profitability still seems far away — and that changes how the entire AI boom gets valued.