Sunday 1 November 2026International EditionRSS
The financial daily of the startup economy
Markets

A $20bn gap in OpenAI's revenue shook AI stocks and exposed how private growth is counted

One newspaper report on October 8 knocked 1.25% off the Nasdaq. The episode shows how much public AI stocks now lean on the unaudited metrics of private companies.

· · 5 min read

On Thursday, October 8, the Financial Times reported that OpenAI had told investors its annualized revenue was approaching $50 billion at the end of September. That was roughly $20 billion below the figure of about $68 billion to $70 billion that had circulated widely at the end of September and had been treated by much of the market as settled fact. Within hours the Nasdaq Composite fell 1.25 per cent, its worst session since mid-August, and the companies whose order books depend on OpenAI's spending took the heaviest losses.

The selloff did not last. By Friday, October 9, the major indexes had recovered, and the S&P 500 finished the week 1.2 per cent higher. But the episode deserves more attention than its brief price impact suggests. It showed that a single unaudited number from a private company can now move hundreds of billions of dollars of listed market value, and that nobody outside a handful of investor meetings knows exactly how that number is built.

What the report said, and what it did not

The distance between $50 billion and $70 billion was not, according to the reporting, a matter of a missed forecast. It was a matter of definition. CNBC, which confirmed the $50 billion figure from an investor presentation, cited a person familiar with the matter who said the larger number included gross revenue from OpenAI's partners, an adjustment intended to make OpenAI comparable with Anthropic. The FT described the higher figure as the product of attempts by OpenAI's own investors to produce a like-for-like comparison with its rival.

The accounting difference is real. Anthropic counts gross sales made through cloud partners such as Amazon. OpenAI counts only its own share, and that share is reduced by the 20 per cent it passes to Microsoft. On those conventions, Anthropic reported an annualized run rate above $65 billion at the end of July, and Reuters, as cited by The Next Web, put Anthropic's second-quarter revenue at $11.5 billion against $6.7 billion for OpenAI.

OpenAI itself did not hide behind the headline. It reported 77 per cent growth in its total run rate during the third quarter and 107 per cent growth in its enterprise business. On Friday, Bloomberg reported that the company expects its annualized revenue to reach or exceed $70 billion by the end of 2026, a forecast shared with investors during talks on a new funding round. In other words, the $70 billion figure was a target, not a September reading.

Who paid for the confusion

The pattern of losses on October 8 was a map of exposure to OpenAI's compute budget. CoreWeave fell about 8 per cent. Oracle, which holds large contracts with OpenAI, lost 5.5 per cent. Intel dropped 5.3 per cent, Super Micro Computer almost 5 per cent, AMD and Broadcom about 4 per cent each, and Nvidia 2.9 per cent. The Nasdaq closed at 27,193.34. The S&P 500 slipped 0.5 per cent to 7,765.36, while the Dow Jones Industrial Average edged up 0.1 per cent to 51,231.64, helped by energy shares as oil climbed.

The asymmetry matters. Investors did not sell the AI complex because demand had fallen; nothing in the report suggested that. They sold because the revenue line that is supposed to justify the largest buildout in the history of the technology industry turned out to be less transparent than assumed. Holger Mueller of Constellation Research told SiliconANGLE that it is still not clear how much funding AI companies will ultimately require, which is the question every supplier's valuation implicitly answers.

The backdrop amplified the reaction. On Wednesday, October 7, the yield on the 10-year Treasury note reached nearly 5.37 per cent, its highest level in 24 years. When the discount rate on long-dated cash flows is that high, the market has less patience for uncertainty about when those cash flows will arrive. The yield eased to about 5.23 per cent on Thursday and stayed below 5.25 per cent on Friday, which gave equities room to recover, but it did not change the underlying arithmetic.

The private valuation question

CNBC put its most recent valuation at $852 billion, following a $122 billion funding round closed in March. The Next Web, citing the latest reporting, said the company is now seeking $30 billion or more at a pre-money valuation of $1.4 trillion, with Abu Dhabi's MGX among the possible investors. OpenAI confidentially filed a draft prospectus in June but has said it will not list this year, pointing to the debate over AI safety, and it is reportedly looking at 2027.

Anthropic is moving faster. Bloomberg has reported that it could go public as soon as November, and it is reported to be seeking a valuation of $2 trillion for its listing. Reuters, citing a leaked prospectus, reported that Anthropic recorded $4.6 billion of revenue in 2025 and a net loss of $42 billion.

These two companies are now being valued on run-rate figures that are defined differently, disclosed selectively and, until a prospectus is made public, unaudited. The October 8 episode is a reminder that a run rate is a snapshot of one month multiplied by twelve. It can be gross or net of partner shares, and it can include or exclude revenue that is about to disappear. The Next Web noted that an earlier figure above $40 billion, cited in mid-August, included Azure revenue that OpenAI stopped receiving around April, when its commercial agreement with Microsoft was revised.

Why private companies should care

The direct lesson is for the AI suppliers and late-stage startups that price themselves against OpenAI and Anthropic. When the reference point moves by $20 billion on a definitional change, every comparable multiple built on it moves too. Secondary market prices for AI shares, internal valuations at crossover funds and the pricing of the next wave of AI listings all depend on these numbers.

The second lesson concerns the public market's role as a check. Before Thursday, the market had absorbed a widely reported figure without asking how it was assembled. It took a newspaper, not a filing, to force the distinction between gross and net. Once Anthropic's prospectus is public, investors will for the first time have audited numbers from one of the two leading model developers, and comparisons will become harder to massage.

Our view

The October 8 selloff was short, but it was not noise. It priced, for the first time in this cycle, the risk that the private AI leaders' headline numbers are softer than the market had assumed, and it did so at a moment when long-term borrowing costs sit at a 24-year high. Founders should take the hint: investors are about to demand that run rates come with definitions attached, and those who report gross figures as if they were net will be marked down when the difference surfaces. For venture investors, the more important date is not OpenAI's eventual listing but Anthropic's, which may arrive within weeks. An audited prospectus will reset the comparables that much of the AI economy is priced against, in one direction or the other. Until then, the honest stance is to treat any run rate without a stated basis as a range, not a number, and to watch the 10-year yield as closely as the revenue headlines.

Sources

Filed under Markets. Written by InvestTimes Editorial under our editorial standards. Commentary and analysis for information only, not investment advice.

More in Markets

All Markets analysis