OpenAI's $50 Billion Run Rate Meets a $70 Billion Target

OpenAI told investors its annualized revenue was roughly $50 billion at the end of September, below the $68 billion figure that had circulated weeks earlier. The company also expects to reach or exceed a $70 billion annualized pace by year-end. Those numbers have different jobs: one describes a reported recent pace, another was an earlier estimate built on a broader accounting basis, and the third is a target for a later date.

Shares of Nvidia, Oracle and CoreWeave fell on October 8 as the lower figure became public. Oracle lost more than 5%, while the Nasdaq was down 1.4% in afternoon trading; rising oil prices and Treasury yields also weighed on stocks. The trading showed how far an OpenAI number can travel beyond OpenAI's own investors. It cannot tell us how much of each stock's decline came from that number.

The immediate question is what the $50 billion measures. OpenAI put that approximate annualized figure before investors. A person familiar with the matter told CNBC that the earlier $68 billion figure included gross revenue from OpenAI's partners so investors could compare it more directly with rival Anthropic. On that account, the apparent drop reflects a change in what was counted, rather than evidence that OpenAI suddenly lost $18 billion of sales. That explanation comes from a person familiar with the matter, not a published reconciliation from the company.

The forward-looking number belongs on a separate line of the ledger. OpenAI expects its annualized revenue to reach or exceed $70 billion by the end of 2026, driven largely by enterprise business. Its September figure was roughly $50 billion. A year-end target can coexist with a lower September reading if sales keep growing. It also asks investors to believe the growth still to come.

What a Run Rate Buys You

Annualized revenue converts a shorter stretch of sales into a yearly pace. A common approach takes one month's revenue and multiplies it by 12, The Information explained. That makes a fast-growing company easier to describe than a completed year's sales would. It also gives the latest month an outsized role: a strong month can lift the headline before a full year of comparable sales exists. The precise period behind OpenAI's September figure, and whether the year-end target uses the same partner-sales basis, have not been made public in these accounts. Neither figure is a statement of completed 2026 revenue or profit.

Investors have a reason to watch the pace. OpenAI's investor presentation highlighted 77% growth in its overall run rate during the third quarter and 107% growth in its enterprise run rate, a person familiar with the presentation said. Enterprise sales are also central to the year-end expectation. If businesses keep paying for OpenAI's products at the implied pace, a September snapshot could become stale quickly. If they do not, annualizing a short period will have made the destination look closer than it was.

The attraction of a current snapshot becomes clearer against OpenAI's history. Sacra estimates that its annualized revenue rose from $36 million in 2021 to $200 million in 2022, $2 billion in 2023, $6 billion in 2024 and $20 billion at the end of 2025. Sacra puts the pace at $40 billion in July 2026 and $50 billion in September. These are estimates and run rates, rather than a series of audited annual sales totals. They describe a business changing fast enough that last year's number may say little about today's scale. They also show why a single recent-period number has become such a tempting shorthand.

The shorthand leaves out the question any business eventually has to answer: what remains after the bills? Sacra describes consumer subscriptions and advertising, business subscriptions, usage-based fees for Codex and developer APIs, and licensing partnerships among OpenAI's revenue streams. Serving those customers has costs. A higher sales pace can improve the economics if revenue grows faster than those costs; the pace alone cannot show whether that is happening. For an investor assessing the company, or a business deciding how much to rely on a supplier, growth and financial durability are related questions, not interchangeable answers.

Why the Anthropic Comparison Got Complicated

The earlier $68 billion estimate has a recognizable purpose. Investors trying to compare OpenAI with Anthropic need to decide whose share of a sale goes into the headline. Sacra describes Anthropic as recording the full amount a customer pays through cloud partners such as AWS and Google Cloud, then treating the partner's share as an expense. For certain sales through partners such as Microsoft, it says OpenAI records only the share it retains.

Picture a customer paying through a cloud platform. Counting the entire customer payment as revenue produces a bigger top line than counting only the AI company's retained share. The partner still receives its portion either way. Moving from one presentation to the other can change the reported comparison even when the underlying customer transaction has not changed. That is the strongest case for why investors constructed a gross-based OpenAI estimate: they wanted figures on a more similar footing.

The exercise has limits. The two companies' partner arrangements and reporting periods need not match, and a larger gross revenue figure does not establish a larger retained share or a better profit margin. Sacra explicitly cautions that the gross-versus-net treatment does not, by itself, indicate a difference in profitability. Using Anthropic's presentation to estimate OpenAI's scale can help frame a comparison; substituting that estimate for OpenAI's own reported run rate creates the confusion investors encountered in October.

OpenAI is seeking more capital while its numbers receive this scrutiny. CNBC reported early-stage discussions of a potential funding round of around $30 billion, with no finalized term sheet, after a $122 billion round closed in March. That gives the accounting question a practical audience. Prospective investors need to know which sales belong in the headline and how long the latest pace has held. Partners and customers have a different concern: whether the company can keep supplying the products on which they are building. Neither group can settle its question by treating a September run rate as a December result.

The next useful OpenAI revenue figure will carry more information if it comes with the period used to calculate it and a consistent explanation of partner sales. Until then, the $50 billion reading describes where the reported pace stood in September. The $70 billion figure describes where OpenAI expects it to go.