OpenAI’s $50 Billion Figure Needs Its Own Accounting

OpenAI told investors its annualized revenue was nearing $50 billion in September, according to the Financial Times report published October 8. The newspaper contrasts that disclosure with a widely reported $70 billion figure. The supplied report summary establishes the source and reference month of the smaller number. It does not establish the original source, measurement date, or precise definition of the larger one.

That missing comparison matters before anyone declares a $20 billion miss. Annualized revenue generally extrapolates a shorter period’s sales over a year; it does not establish what a company booked during a completed year. Neither figure should be treated here as audited annual revenue. A run rate also needs its own instructions: which period was multiplied, which sales were included, and whether the comparison uses the same method. Annualized revenue has a calendar preference. Investors need to know which one.

Two explanations worth testing

Granted, different measurement periods could explain part of the discrepancy. A September pace could sit below an expectation for a later date without contradicting it. A monthly calculation could also differ from one based on a quarter. These are plausible explanations, not established facts about OpenAI’s accounting. The strongest benign case is that public discussion has placed two differently labeled numbers in the same column and congratulated itself on subtraction.

Still, a mismatch in reporting is not the only possibility. If both figures describe comparable annualized sales at comparable dates, the lower disclosure would raise a substantive question about commercial momentum. Even then, weaker performance against a reported expectation would not automatically mean OpenAI missed formal company guidance. The provenance of the $70 billion figure determines whether investors are evaluating a company commitment, an outside estimate, or something else.

The consequences follow the explanation. A measurement mismatch would call for clearer communication and recalibrated comparisons. A genuinely weaker revenue pace could make financing assumptions less comfortable, particularly where those assumptions depend on rapid sales growth. But revenue alone would not establish profitability, cash available for investment, or the ability to meet future obligations. A large sales number cannot answer every question merely by being large.

OpenAI can narrow the uncertainty by disclosing the calculation behind the September figure, comparable earlier periods, and the status of any forward revenue expectation. The $70 billion figure needs an attributable source and a reference date as well. If comparable disclosures show sustained growth consistent with the earlier expectation’s timing, the reporting-mismatch explanation strengthens. If they show a persistent shortfall against a documented company forecast, the weaker-performance explanation strengthens. Until then, the gap is a question worth investigating, not an accounting verdict.