The company that convinced the world AI is worth trillions just admitted it's worth 29% less than we thought it was two weeks ago.
The Summary
- OpenAI's annualized revenue is now pegged at roughly $50 billion, not the $70 billion figure that's been circulating in media and pitch decks
- The $20 billion gap represents nearly 30% of the previously reported revenue run rate, a rounding error that would sink most public companies
- This is still growth from the prior year, but the revision raises questions about how AI valuations get set when the underlying numbers are this slippery
The Signal
OpenAI is still on track to generate about $50 billion in annualized revenue, according to people familiar with the company's performance. That's an impressive number for a company that was doing maybe $2 billion in revenue just two years ago. But it's also $20 billion short of the $70 billion figure that's been floating around investor presentations, media coverage, and probably a few term sheets.
The $70 billion number wasn't a wild guess. It was widely reported as OpenAI's current run rate. Now we're learning it was off by 29%. In traditional finance, that's the kind of miss that gets CFOs fired and triggers SEC investigations. In AI land, it's apparently just a revision.
"The $20 billion gap represents nearly 30% of the previously reported revenue run rate, a rounding error that would sink most public companies."
Here's what makes this messy: annualized revenue isn't actual revenue. It's a projection based on recent performance, usually the last month or quarter multiplied out. Companies use it to show momentum, especially in high-growth phases. But it's inherently squishy. A good month gets extrapolated into a $70 billion year. A realistic quarter becomes $50 billion. Neither number is a lie, exactly, but neither is quite the truth either.
The real question is what changed. Did OpenAI's growth slow? Did enterprise customers churn? Did the math just get more honest? Bloomberg notes this is still an uptick from the prior year, which suggests the company is still growing. But if the trajectory flattened enough to lop $20 billion off the run rate, that's a signal investors should care about.
Key factors that could explain the gap:
- Slower enterprise adoption than projected in Q3
- Higher churn among consumer ChatGPT Plus subscribers
- API revenue growing but at lower margins than assumed
- More conservative accounting ahead of a potential public offering
This matters because OpenAI isn't just another startup. It's the flagship for the entire AI investment thesis. When its numbers shift by $20 billion, every valuation model downstream shifts with it. If OpenAI is worth 29% less than we thought, what does that mean for Anthropic, Cohere, and the hundred other foundation model companies raising at nosebleed multiples?
The Implication
If you're in the agent economy, this is a reminder that revenue projections in AI are still more art than science. The models are new, the use cases are evolving, and nobody really knows what sustainable adoption looks like at scale. A $20 billion revision isn't a scandal, but it's a sign that the market is still figuring out what this technology is actually worth.
Watch how this plays in the next fundraising cycle. If OpenAI's numbers are softer than expected, expect investors to start asking harder questions about burn rates, unit economics, and actual customer retention. The era of "just build the best model and revenue will follow" might be ending faster than anyone expected.