Where we heard itBloomberg Daybreak: Europe on Bloomberg Television, Oct. 9, 2026. Host Stephen Carroll with Bloomberg News reporter Anthony Stephens in Hong Kong.
Investors disagreed Thursday on whether OpenAI was reaching targets or falling short. The dispute centered on a definition: what counts as revenue. Reports that OpenAI’s annualized revenue fell well below earlier signals spread quickly, and AI stocks sank on the report.
On Bloomberg television, host Stephen Carroll framed the story: “Bloomberg understands OpenAI is expecting to reach or exceed $70 billion in annualized revenue by the end of the year, driven largely by growth in its enterprise business.” Bloomberg reporter Anthony Stephens then explained that the true net figure for OpenAI coming into September was around $50 billion. OpenAI is privately held, and these numbers have not been independently verified.
Two Defensible Numbers for One Company
Stephens explained, listing MicrosoftMSFT and AmazonAMZN: “This whole issue speaks to kind of a mathematical granularity in what annualized revenue means to the different labs. So we have Anthropic reporting gross numbers and OpenAI net, and what net means is net of all the money that you have to pay to Microsoft or Amazon or all the routes through which these LLMs get business.” Coming into September, he said, “it looks like the true number is around 50 billion.”
An annualized figure, or run rate, projects recent sales across a full year. Gross run rate counts money from customers. Net deducts what OpenAI pays those cloud providers for infrastructure. Both describe the same business; only net shows how much the company keeps, essential for any profitability discussion.
CNBC confirmed the roughly $50 billion end-of-September figure and reported that the $68 billion number widely cited in late September included gross revenue from OpenAI’s partners.
This is ordinary for private company reporting. OpenAI has no obligation to standardize, so two valid figures coexist depending on who measures and why.
Enterprise Demand Sits at the Center of the Outlook
Stephens presented business customers as the foundation of the year-end target. New reporting suggests that “by the end of the year, we will still be at a 70 billion run rate, and that is mostly because it’s going to be driven by enterprise AI demand,” he said, a trend coming through as more companies “use agentic AI in particular, which has a higher revenue potential.” Agentic AI refers to software that carries out multistep tasks on a user’s behalf.
How the market swung on the revenue reports
How a Private Company Moved Public Markets in One Day
Technology shares fell Thursday as confusion spread. Sentiment improved as the gross versus net distinction became clearer, and CNBC reported Friday that “stock futures rise as tech shares rebound following sell-off.”
The Bloomberg segment also flagged international effects: a materially lower move in SoftBank shares in Asia, and the Nvidia-backed data center company Firmus pulling its Australian listing, with plans to raise private capital and pursue a US listing later.
An unlisted company that files no quarterly reports and releases no standardized revenue definition triggered a one-day selloff in public technology shares because investors could not agree on what its headline number meant.
Vague Metrics, Real Money at Stake
Stephens closed with the broader issue: “There’s a lot of money at play here, some of these metrics are a bit vague and hopefully we will see more clarity from both Anthropic and OpenAI going forward on kind of the raw metrics around these revenue models.”
Public companies report on a standardized basis so investors can compare businesses. Microsoft and Amazon file with the Securities and Exchange Commission quarterly using defined accounting rules. The largest AI developers are private and do not. As public market valuations lean more heavily on private AI company performance, that gap matters more. Bloomberg’s coverage shows how often these private figures now move the conversation around listed technology stocks.
When a private AI company’s revenue figure moves markets, the first question to ask is which figure it is, and how it was measured.
Anthony Stephens traces the OpenAI revenue confusion to definitions: OpenAI reports net of what it pays partners such as Microsoft and Amazon, while, he said, Anthropic reports gross. Bloomberg reports OpenAI expects to reach or exceed $70 billion in annualized revenue by year-end, up from about $50 billion at the end of September. The risk is that a private company with no standard disclosure can still swing public tech stocks, as Thursday’s selloff showed.