OpenAI’s $20 Billion Revenue Gap Sends Chip Stocks Tumbling as the AI Trade Wobbles

OpenAI told investors its annualised revenue was approaching 50 billion US dollars at the end of September — about 20 billion dollars less than the figure investors had previously been led to expect — according to a Financial Times report published on Thursday, and the revelation sent technology shares sharply lower on both sides of the Atlantic trading day.

The Financial Times, citing financial documents shared with OpenAI’s backers, reported that the ChatGPT maker put the figure at close to 50 billion dollars. That stands in stark contrast to media reports in late September that had pegged the number at roughly 70 billion dollars, based on information shared with investors.

The discrepancy, according to the Financial Times, stemmed from an attempt by OpenAI’s own investors to produce a revenue figure that could be compared directly with the annualised revenue of its arch-rival Anthropic. The crucial difference is in the accounting: Anthropic counts sales made through its cloud partners — Amazon Web Services and Google Cloud — in its run-rate metric, while OpenAI does not, the publication reported, citing a person familiar with the matter. A source familiar with the documents told CNN that the 70-billion-dollar figure did not originate from OpenAI at all, and had likely arisen from a desire by some firms to measure the company against Anthropic’s gross revenue, which includes cloud-provider sales. OpenAI’s figure, by contrast, is based on net revenue. OpenAI declined to comment, according to the Financial Times and CNN.

The market reaction was swift and unforgiving. The Nasdaq Composite sank 1.25 per cent on Thursday, its worst day since mid-August, according to CNN, with the declines accelerating after the Financial Times report was published in the middle of the session. The Nasdaq 100 fell 1.4 per cent, its poorest session in seven weeks, and the S&P 500 slipped 0.5 per cent, according to market data cited by MarketWatch and Australian market wrap coverage, while the Dow Jones Industrial Average edged 0.1 per cent higher. The Philadelphia Semiconductor Index tumbled 3.4 per cent, and the S&P 500’s information technology sector dropped 1.8 per cent — its largest one-day fall since mid-September, according to MarketWatch.

Selling spread across the full supply chain of the artificial intelligence buildout. Nvidia fell 2.9 per cent and Intel slid 5.3 per cent, according to CNN, while AMD dropped 3.9 per cent, Broadcom — which is designing a custom chip with OpenAI — lost 4.3 per cent, and memory makers Micron and Sandisk fell 4.8 and 4.9 per cent respectively, according to MarketWatch. Oracle, which sits at the centre of OpenAI’s data-centre expansion, dropped roughly five and a half per cent, and Microsoft, a key OpenAI backer, slipped into the red, losing 1.2 per cent at midday to about 523 dollars, according to Dow Jones Newswires. Taiwan Semiconductor Manufacturing’s US-listed shares fell 3.3 per cent to about 457 dollars, and Celestica, a chip supplier, slid 5.9 per cent, the newswire reported.

Even companies delivering spectacular results could not escape the downdraught. Samsung Electronics said it expected this quarter’s operating profit to more than double what it earned across all of 2025 — yet its shares still fell 2.4 per cent, a measure of how elevated expectations have become for AI beneficiaries, according to The Wall Street Journal. Taiwan Semiconductor Manufacturing reported that its September sales had risen 55 per cent year-on-year, but the figure was down 0.6 per cent on the previous month and its US-listed shares slipped 3 per cent, the Journal reported.

Ross Mayfield, an investment strategist at Baird, told CNN the episode could reverberate far beyond one company’s accounts. “There are going to be tremors throughout all of the related sub-industries,” he said, adding that investors’ entire outlook for technology stocks rests on the conviction that demand for AI will keep growing — and that any wrinkle in that story “could send ripples throughout the supply chain for the AI buildout.”

The timing of the report could hardly be more charged. OpenAI is in negotiations over a fresh private fundraising round that could value the company at roughly 1.4 trillion dollars, according to the Financial Times. The company confidentially submitted paperwork for an initial public offering in June, but the listing — once widely expected for this autumn — has been pushed back to early 2027, according to TechCrunch. Anthropic, meanwhile, is expected to list on public markets as early as November. The Financial Times has described the 20-billion-dollar gap as “a massive gap likely to damp optimism about the growth of AI demand.”

OpenAI’s finances have been under scrutiny all year. TechCrunch noted that the company raised 122 billion dollars in a March funding round alone, and that its leaked 2025 financials showed it generated about 13 billion dollars in revenue while spending significantly more. Shay Boloor, chief market strategist at Futurum Equities, offered a sharper interpretation of Thursday’s numbers in emailed comments to Investopedia: “The $20B gap between OpenAI’s real run-rate and reported number is simply the value of sales flowing through [AWS], Google and Microsoft Azure, but it’s another great reminder of the winner of the AI model race keeps looking like whoever owns the distribution toll booth, which is Amazon, Google, Microsoft, Meta and Palantir.”

The Financial Times report also lands against a backdrop of mounting unease about how the AI buildout is being financed. In recent weeks, reports detailed the latest debt financing arrangements between Broadcom and OpenAI, following revelations that SpaceX and Nvidia were arranging chip finance of up to 40 billion dollars, and that Broadcom and Anthropic were discussing debt financing on a similar scale, according to Reuters. The growing reliance on debt — and what some analysts describe as vendor financing, in which suppliers effectively fund their own customers’ purchases — has raised the stakes for the earnings season now getting under way, with investors watching for evidence that AI demand can justify the capital expenditure.

Analysis: Why It Matters

This story is not really about a number. It is about the fragility of the assumptions on which a historic market rally is built.

Begin with the metric itself. “Annualised revenue” — sometimes called run-rate revenue — is a creature of the private startup world. It typically takes a single month’s sales and multiplies by twelve, a shortcut that flatters any fast-growing company and can blur important differences in pricing, costs and contract quality. AFP described the measure as one that is “sometimes criticized as imprecise or even misleading.” When a private company worth a putative 1.4 trillion dollars is valued in large part on such a number, the precision of that number matters enormously — and Thursday’s gap shows just how squishy it can be. One month’s gross cloud-partner billings can be 20 billion dollars away from another month’s net recognised revenue, depending on who is counting and why.

Second, the gap illuminates the real economics of the AI boom. The most quotable line of the day belonged to Shay Boloor of Futurum Equities, who pointed out that the missing 20 billion dollars is essentially the value of sales flowing through the cloud platforms of Amazon, Google and Microsoft — “another great reminder,” he said, that the winners of the AI model race increasingly look like whoever owns the distribution toll booth. That observation cuts to the heart of the debate raging across Wall Street: the model-makers spend billions building frontier systems, but the cloud providers sit on the transaction, take their margin, and increasingly sell their own competing models. If investors are asked to pay trillion-dollar valuations for AI labs, they need to know whether the cash registers are the labs’ or their landlords’. Thursday’s answer was uncomfortably ambiguous.

Third, the episode exposes a valuation gap that goes far beyond OpenAI. Nvidia, Broadcom, AMD, Intel and the memory makers have enjoyed one of the most powerful rallies in market history on the promise that AI demand is effectively insatiable. But that demand chain runs through a handful of private labs whose finances are opaque and whose spending is increasingly financed by debt rather than cash flow. Reuters has documented a wave of debt financing tying chipmakers directly to their customers — Broadcom to OpenAI, Nvidia to SpaceX, Broadcom to Anthropic. When a supplier lends to its customer so the customer can buy the supplier’s product, the resulting revenue is real, but the quality of that demand is a fair question. The Financial Times’ phrase — “a massive gap likely to damp optimism about the growth of AI demand” — is the polite version of that concern.

Fourth, consider the sequencing risk. Anthropic is expected to come to market as early as next month, reportedly on track to reach 100 billion dollars in annualised revenue this year, according to the Financial Times via Investopedia. OpenAI, the larger and better-known name, has just delayed its own listing to early 2027 and is now negotiating a private round at a valuation that would make it one of the most valuable companies on earth. If Anthropic lists first — with the larger revenue number and the cleaner growth story — it could set the pricing benchmark for the whole sector, and Thursday’s selloff suggests the market is already recalibrating.

Finally, there is the interest-rate environment, which is doing the AI trade no favours. The 10-year Treasury yield eased to about 5.22 per cent on Thursday after a solid 30-year bond auction, but it remains near multi-decade highs, and the latest Federal Reserve minutes showed most policymakers believed another rate increase would likely be appropriate by the end of the year, according to Reuters. High long-term yields mechanically compress the multiples investors will pay for future earnings — precisely the kind of earnings the AI trade is built on. Technology stocks have been the market’s ballast in recent weeks precisely because everything else has been struggling under the weight of borrowing costs. When the ballast itself springs a leak, there is not much left to hold the rally up.

None of this means the AI buildout is a fiction. Taiwan Semiconductor Manufacturing’s 55-per-cent sales jump and Samsung’s profit expectations are reminders that real silicon is being bought in real volumes. But markets price narratives, and Thursday’s narrative changed: from “AI revenue is compounding faster than anyone predicted” to “nobody is quite sure whose revenue it is.” For a sector that has been priced for perfection, that is a painful transition.

What to watch now: the earnings season about to begin, when the hyperscalers’ capital-expenditure commentary will either validate the spending or deepen the doubts; Anthropic’s expected listing next month, which will put a public-market price on AI-lab revenue for the first time; the terms of OpenAI’s fundraising round, which will show whether private investors still believe the 1.4-trillion-dollar story; and Microsoft’s results in particular, given its role as OpenAI’s biggest backer and its own heavy AI spending. If revenue growth re-accelerates in the next quarter, Thursday will be remembered as a one-day air pocket. If it does not, the tremors Ross Mayfield warned of may only just be starting.

Sources

• Financial Times via TechCrunch: OpenAI’s revenue is reportedly $20 billion less than previously projected
• CNN: Tech stocks drop after report that OpenAI’s revenue is lower than expected
• MarketWatch: Micron, Nvidia and AI chip stocks fall as report on OpenAI’s revenue causes ‘undue concern’
• Dow Jones Newswires via Morningstar: Oracle, AMD, Others Slide After Report of Revision to OpenAI Revenue Estimate