The number landed Thursday and sent a jolt through the AI trade. OpenAI told investors its annualized revenue was approaching $50 billion at the end of September, about $20 billion less than the $70 billion figure that had been circulating among investors for weeks.
Within hours, the Philadelphia Semiconductor Index fell 3.4%. Nvidia dropped about 3%, Oracle nearly 6% at one point, and CoreWeave almost 8%. Broadcom fell more than 4% intraday, and memory makers Micron and SK Hynix slid with it. The Nasdaq 100 fell 1.4%.
The selloff had a strange trigger, because the gap did not reflect lost sales. The $70 billion number was a figure built by investors trying to compare OpenAI directly with its rival Anthropic, according to the Financial Times, which first reported the $50 billion figure. The two companies count revenue differently.
Anthropic records the full value of sales made through cloud partners such as Amazon Web Services and Google Cloud, then logs the cloud provider’s cut as an expense. OpenAI books only its own share of proceeds from partner arrangements, including those with Microsoft. Both approaches comply with accounting standards; the difference turns on who controls the customer relationship.
So the apparent shortfall was mostly a matter of arithmetic. OpenAI told employees the same day that it still expects annualized revenue to reach $70 billion by the end of the year.
The company has been growing fast by any ordinary measure. Its annualized revenue stood at around $40 billion in August, and it has told investors to expect growth of more than 70% through the end of the year. In the third quarter it posted total run-rate growth of 77%, and its enterprise business grew at a 107% run rate, according to a person familiar with the matter.
The revenue question matters because OpenAI is also a heavy spender and a heavy raiser. It is in the market for tens of billions of dollars in fresh capital to fund the data centers and chips its models require, and its valuation depends on investors believing the revenue curve will keep bending upward. A number that looks smaller than expected, even for reasons of accounting, complicates that pitch.
The market’s reaction said more about investor mood than about OpenAI’s books. AI chip stocks have been priced for perfection after a long climb, and any whiff of weakness in the sector’s most important customer is enough to trigger selling. OpenAI is the flagship buyer of the chips, data centers and power on which the entire AI infrastructure trade is built.
The confusion is a symptom of how little the public knows about the finances of the industry’s most valuable private company. OpenAI releases no financial statements. Its known figures are fragments: about $13 billion in revenue in 2025, a $122 billion funding round in March, and an IPO that has been pushed to early 2027.
Analysts called the selloff overdone. The gap, several said, was a clarification of how revenue is reported rather than a signal about underlying demand. One noted that the reaction showed how tightly the market now tracks the two largest US AI companies, and how little it has to go on beyond leaked numbers.
The episode also exposed a fragile habit. Investors, starved for official data, have been extrapolating from whatever figure leaks out of the AI labs, then repricing entire sectors when a new figure arrives. A market that moves billions of dollars on an accounting distinction is a market flying partly blind.
The two firms’ methods are both defensible, but the difference is not trivial. Anthropic’s approach makes its revenue look larger, which helps when it touts momentum. OpenAI’s approach makes its revenue look smaller, which helps when it argues it is still early in monetizing a much larger opportunity.
OpenAI has reason to be careful about the number it puts in front of investors. The company is in the middle of raising tens of billions of dollars at a valuation that depends on a steep revenue curve. Understating near-term revenue, then beating it, is easier to defend than the reverse.
For the chipmakers caught in the downdraft, the episode is a warning about concentration. Nvidia, Broadcom, Micron and the rest have ridden a wave of spending from a small number of AI labs. When one of those labs sneezes, the whole supply chain reaches for a tissue.
The gap between the two figures, in other words, is not a correction of fact so much as a correction of comparison. Investors had been comparing OpenAI’s number to Anthropic’s as if the two were measured the same way. They are not, and the market spent Thursday discovering the difference at a cost of billions of dollars.
OpenAI declined to comment. The $50 billion figure, lower on paper and higher in credibility, is now the one investors will anchor on – until the next leak sends the market reaching for the same tissue again.
