How OpenAI counts revenue makes its $70B run rate unclear

OpenAI’s annualized revenue was roughly $50 billion at the end of September, according to people familiar with figures the company shared with investors. It expects that figure to reach or exceed $70 billion by the end of 2026, with growth in its business and enterprise sales doing most of the work.

A run rate, or annualized revenue, is a projection. You take sales from a recent short stretch, such as the last month, and multiply it out to a full year. It shows momentum, but it is not the same as money already collected over 12 months.

OpenAI also told investors its total run rate grew 77% in the third quarter, and its enterprise run rate grew 107%, according to a person familiar with the presentation. OpenAI declined to comment publicly on any of the figures.

The confusion built up over two months. In mid-August, reports put OpenAI’s run rate past $40 billion. That figure later turned out to include money OpenAI used to get when Microsoft sold OpenAI models to its Azure cloud customers, revenue that stopped around April when the two companies updated their deal.

On Sept. 29, reports said OpenAI was nearing $70 billion a year. That number was repeated widely. On Oct. 8, a financial newspaper reported that the $70 billion did not come from OpenAI at all. It came from OpenAI’s own investors, who were trying to recalculate the company’s numbers the way rival Anthropic counts its revenue, so the two could be compared directly.

Later that day, another report said the earlier figure traced back to a single investor’s estimate. That estimate followed a comment from OpenAI’s chief financial officer, Sarah Friar, that the run rate grew 70% between the end of June and the end of September.

The heart of the problem is a simple accounting choice. When Amazon, Microsoft or Google sell Anthropic’s Claude models to their own cloud customers, Anthropic counts the full value of those sales. That is called gross revenue.

OpenAI does it the other way. When partners such as Microsoft sell its models, OpenAI counts only the share it actually receives, which is net revenue. Its figures have generally included the 20% of revenue it shares with Microsoft.

Neither method is a trick, but they produce very different totals. Anthropic’s annualized revenue passed $65 billion at the end of July. In the second quarter, Anthropic’s quarterly revenue topped OpenAI’s for the first time, $11.5 billion against $6.7 billion, according to a report citing the companies’ figures. Comparing the two is like comparing a store that counts every sale at its franchise locations with one that counts only the fees it collects from them.

If you use ChatGPT, nothing changes today. The service is not shrinking, and both figures point to a business that is still growing fast. The bigger effect is on money.

On Oct. 8 the Nasdaq Composite fell 1.25%, its worst day since mid-August. Oracle, which has multibillion-dollar contracts with OpenAI, dropped 5.5%. Intel fell 5.3%, Nvidia 2.9% and CoreWeave, a company that rents out AI computing power, 8%. If you own an index fund or a retirement plan heavy in tech, you felt some of that.

The episode also shows how much of the AI boom rests on numbers that are not audited or published. Private companies share run rates selectively, often during fundraising. Until they file public financial statements, treat any single revenue headline with care, especially when it comes from people “familiar with the matter” rather than the company itself.

OpenAI shared its forecast while raising new money. It wants to raise at least $30 billion, with the company valued at $1.4 trillion before any of that cash is counted. Investment funds from the United Arab Emirates, including Abu Dhabi’s MGX, have discussed putting in up to $10 billion together, and the asset manager BlackRock has been in talks to join.

For context, OpenAI raised $122 billion in March at an $852 billion valuation. It has pushed a stock market listing back to at least 2027, citing its focus on AI safety. Anthropic, by contrast, is expected to go public as soon as November, and the two companies’ numbers will be compared even more closely once one of them has to publish audited results.

To get from about $50 billion to $70 billion by December, OpenAI needs to add roughly $20 billion in run rate in a single quarter. Given the 77% jump it reported for the third quarter, investors seem willing to believe it. The next real test comes when the funding round closes and its final terms become public.