OpenAI's revenue is growing fast—but maybe not as fast as some headline numbers suggested. A Financial Times report says the AI lab told investors its annualized “run-rate” revenue was nearing $50 billion at the end of September, not the $70 billion figure that had been floating around in market chatter.
The discrepancy isn't a sign that demand is collapsing. Instead, it comes down to a quieter, more technical issue: how each company counts revenue when sales flow through cloud partners like Amazon Web Services and Google Cloud.
What is run-rate revenue, anyway?
“Run rate” is a common shortcut in the tech world. It takes a recent month's revenue and multiplies it by 12 to estimate what a company would earn in a full year if that pace held. For fast-growing companies, that can exaggerate scale—especially if growth is accelerating or if the number includes sales that aren't directly recognized by the company.
According to the FT, OpenAI and Anthropic—two of the biggest AI model developers—both sell access to their technology directly and through major cloud platforms. The report says Anthropic counts cloud-partner sales through AWS and Google Cloud in its run-rate figure, while OpenAI does not. That means the two headline numbers aren't measuring the same thing, even though they look comparable at a glance.
Think of it like a restaurant's revenue: one owner might count every dollar that comes through the door, including catering orders booked through a third-party app. Another might only count what the restaurant itself bills directly. Both can honestly say they're reporting revenue, but the numbers tell different stories about the business.
Why the gap matters for investors
Run-rate revenue is often used as a shorthand for valuing buzzy private companies, especially when investors talk in terms of enterprise value-to-sales—a simple “price tag divided by revenue” multiple. If OpenAI's figure excludes cloud-partner sales that a peer includes, the multiple you'd back into can look meaningfully different without any change in actual demand.
That can ripple through private-market valuations, secondary trading, and eventually IPO pricing if either company goes public. As we've seen in recent market reactions to AI revenue reports, investors are paying close attention to how AI companies turn hype into hard numbers.
The episode also highlights a broader point: private companies aren't held to the same standardized accounting rules as public ones. Until a company files for an IPO, its financials can be presented in ways that are hard to compare across peers. That's not necessarily deceptive—it's just that the rules are looser.
What happens next
Reuters said it couldn't independently verify the FT report and that OpenAI didn't immediately respond to a request for comment. Still, the episode is a reminder that private-market metrics can be squishy until companies face public-style disclosures.
If OpenAI or Anthropic eventually file to go public, standardized accounting and prospectus-level detail should make their growth, margins, and go-to-market strategies easier to compare. That could shift expectations in IPO pricing and pre-IPO secondary trading, as investors get a clearer picture of what's actually being counted.
For everyday investors, the takeaway is simpler: when you see a big revenue number for a private company, ask what's included. Run-rate revenue only works if everyone's counting the same dollars. Until then, treat headline figures as directional, not definitive.
As the AI sector matures, expect more scrutiny on how companies define their metrics—and more stories like this one that peel back the curtain on the math behind the headlines. For a deeper dive into why $50 billion and $70 billion can both be true, check out our explainer on OpenAI's revenue math.
