OpenAI has told investors its annualized revenue is approaching $50 billion, according to Financial Times reporting summarized in TechCrunch’s October 8 account. That is roughly $20 billion below the estimate circulated in earlier reporting, a substantial difference for investors assessing the company’s scale and its competition with Anthropic.
According to the report, the earlier $70 billion figure emerged from attempts by OpenAI’s investors to make a direct comparison with Anthropic’s annualized revenue. The companies reportedly count sales made through cloud partners differently: Anthropic includes those sales in its annualized figure, while OpenAI does not.
The revision depends on how revenue is measured. It is not an audited disclosure showing that OpenAI lost $20 billion in business, and neither headline figure establishes how much revenue OpenAI earned over a completed financial year.
The lower estimate changes the investment discussion, but it cannot explain whether underlying demand changed, how much either company retains from partner sales, or which measure provides the fairest comparison.
The $70 Billion Figure Was a Comparison Exercise
TechCrunch traces the higher estimate to information shared with OpenAI investors and their efforts to produce a figure comparable with Anthropic’s reported run rate. It describes the newer figure, approaching $50 billion, as the annualized revenue OpenAI has told investors it is generating.
Headlines can blur the distinction between a company’s reported operating metric and an investor-constructed comparison. Adjusting a figure can serve a legitimate analytical purpose: when two businesses measure sales differently, investors may try to put them on a common basis before comparing their size. Readers still need to know what was adjusted, why, and whether the result represents the metric the company itself reports.
The available account provides no reconciliation showing how the approximately $50 billion figure became approximately $70 billion. It identifies investor comparisons and the different treatment of cloud-partner sales, without establishing that the entire difference can be assigned to one disclosed adjustment.
The reporting supports a limited conclusion: the widely circulated estimate and the figure reportedly communicated by the company were different. It demonstrates neither a sudden revenue collapse nor that the higher number was necessarily a valid, fully comparable measure of OpenAI’s business.
Cloud-Partner Sales Complicate the Anthropic Comparison
The central methodological issue is cloud-partner sales. TechCrunch says Anthropic counts these sales in its annualized revenue figure and OpenAI does not.
A measure that includes partner-mediated sales may capture commercial activity beyond the scope of one that excludes them. Comparing those totals without accounting for that boundary risks treating different categories of activity as equivalent.
Several questions sit behind the headline numbers:
- How much does the end customer spend?
- How much does the cloud partner record?
- How much does the model provider receive?
- Which of those amounts enters the annualized figure being discussed?
The available report does not answer those questions. It establishes only that the inclusion rules differ, and supplies too little detail to determine how either company’s treatment maps onto formal revenue recognition. It would be premature to label one total “gross revenue” and the other “net revenue,” or to assume that a particular share of cloud spending flows directly to either lab.
The correction therefore cannot establish a clean revenue ranking between OpenAI and Anthropic. OpenAI’s company-reported run rate being lower than an earlier estimate does not resolve how it compares with a rival whose metric includes a different sales category.
Investors would need figures calculated on a common basis, or enough disclosure to construct that basis themselves. Placing two annualized totals next to each other does not make them comparable.
Annualized Revenue Is Not a Completed Year’s Revenue
An annualized run rate projects revenue from a shorter measurement period across a year. It describes the scale implied by a recent pace of business, rather than adding up revenue already earned during a completed twelve-month period.
For a growing business, the difference can be substantial. A recent period may reflect more activity than earlier months, producing an annualized number above the revenue recorded over the preceding year. Conversely, the pace used in the calculation might not continue.
The reporting does not identify the precise measurement window behind the approaching-$50-billion figure. Readers should not assume it represents a particular month, quarter, or trailing twelve-month period.
Even an audit would leave the distinction between a historical total and a forward extrapolation intact. The key questions remain what the metric measures, which period supports it, and what assumptions are required to extend it across a year.
Here, the figures are reported annualized estimates, not audited full-year revenue disclosures. They establish neither profitability nor cash generation, and they do not reveal the costs required to deliver the underlying AI services. A run rate can describe commercial scale without independently showing how durable that scale is or how much economic value the company captures.
A Lower Revenue Denominator Changes the Investment Case
The roughly $20 billion difference matters even if it reflects measurement rather than a deterioration in sales.
Investors often assess a company’s valuation relative to revenue. Holding valuation constant, a smaller revenue figure produces a higher revenue multiple. Using the rounded headline numbers, dividing the same valuation by $50 billion rather than $70 billion raises that multiple by 40%.
This is an illustrative calculation, not a statement about OpenAI’s actual valuation. It shows how an investment argument built around the higher estimate may look less attractive when evaluated against the lower one.
There is a risk in dismissing the adjustment, too. If investors deliberately adjusted OpenAI’s figure to match Anthropic’s reporting scope, rejecting it without examining its construction could produce another misleading comparison. The question is whether the adjustment is transparent and economically justified, regardless of whether it makes the number larger.
Different measures can answer different questions. A broad measure of customer spending may help assess market reach; revenue attributable to the model provider may be more useful when examining its own financial performance. Neither should silently substitute for the other.
Sources
- TechCrunch’s October 8 accounttechcrunch.com
