The Great AI Revenue Mirage

Illustration: a glittering mirage of a futuristic city floating above a desert of burning cash, flat vector, warm reds and oranges

October 2026. The leaves are turning, pumpkin spice is flowing, and the air is thick with the scent of… well, not quite panic, but definitely a faint whiff of "oh crap, maybe this isn't going straight to the moon." The AI bubble — the trade that could only go up — just had its worst wobble in weeks, and for once the numbers, not the vibes, are driving the conversation.

Start with the revenue mirage. Reporting based on OpenAI's own investor materials (summarized in this Medium breakdown) puts the company's annualized revenue around $50 billion — impressive, until you remember the number previously floated to investors was closer to $70 billion. A $20 billion gap between the story and the spreadsheet is not a rounding error. It is the sound of a narrative being repriced in real time.

The market noticed. On October 9, the Nasdaq 100 dropped 1.4% — its worst one-day fall in seven weeks — while the Philadelphia Semiconductor Index plunged 3.4%, as Moneyweb's market wrap reported. One bad day proves nothing, but Bank of America's derivatives team gave the wobble a name worth remembering: October 8 was, in their words, the "best rehearsal for an AI unwind" — stocks and yields tumbling together, the exact pattern you'd expect if the AI trade ever really cracked, as ZeroHedge relayed. The transmission chain matters here: stress starts in the private AI ecosystem — the rounds, the commitments, the circular deals — then hits the Nasdaq, then drags the bond market into safe-haven mode. Public investors are the last to know and the first to pay.

Now the actual numbers, because those are always fun, especially when they involve more red than a stop sign. Per Quartz's reporting, OpenAI lost a staggering $38.5 billion in 2025 on revenue of $13.07 billion. Let's be real: if your local bakery lost $38.5 billion, they'd be selling the sourdough starter on eBay, not raising another round. This isn't just burning cash; it's setting fire to a vault full of money while wearing a suit made of hundred-dollar bills. And yet the valuation kept climbing — CNBC reported an $852 billion figure, an IPO prospectus filed in June 2026, and talks for yet another $30 billion round — all based on a future that seems perpetually just around the corner, like that diet you're definitely starting next Monday.

Then there's Anthropic, another poster child for boundless potential. Reported figures put its compute commitments at a mind-boggling $518 billion over the coming decade. Its 2025 revenue? A comparatively tiny ~$4.6 billion. That's like buying a private jet for every employee when your company's main product is artisanal paper clips. Who needs profits when you have potential? Who needs a sustainable business model when the next breakthrough is always one more server rack away? It's the classic "we'll make it up in volume" — where the volume is compute, and the making-it-up is doing a lot of heavy lifting.

None of this means AI is fake, or that the technology stalls here. It means the financial story and the technology story have come unglued. The models keep getting better on a curve anyone can measure. The balance sheets keep getting worse on a curve anyone can also measure — if they bother to look. The whole situation feels less like innovation and more like a high-stakes game of musical chairs, where the music is venture capital flowing and everyone hopes they're not the one standing when the funding dries up. The real winners, as always, are the early investors and founders who've already cashed out. For everyone else, it's a front-row seat to what BofA so optimistically called a rehearsal.

So as the days get shorter and the market gets choppier, maybe it's time to put down the AI-generated stock tips and take a hard look at the actual numbers. The future of AI might genuinely be bright. The present financial reality for some of its biggest players looks suspiciously like a very expensive hole in the ground. And if this was just the rehearsal, brace yourself for the main act.

To be fair, there is a version of this story where the bulls win. If enterprise AI spending keeps compounding at its current clip, if inference costs keep collapsing faster than usage grows, and if the $50 billion revenue line turns out to be the floor rather than the ceiling, then today's losses read, in hindsight, like the cost of buying a monopoly early. Amazon lost money for years, too. The difference is that Amazon's losses bought warehouses and logistics — assets that got cheaper to run every year. AI's losses buy compute commitments that expire whether the revenue shows up or not. That asymmetry is the whole argument, and right now the bears have the arithmetic.

Revised and rerun from the October 17 draft. Figures are as reported by the linked sources; revenue and loss numbers for private companies are estimates, not audited financials.