Sources familiar with the financials tell Axios the annualized run rate has grown more than 70% since the start of Q3 — but the cost side of the ledger remains entirely unseen.
This is sourced reporting, not a company disclosure. Axios attributes the figures to “sources familiar with the financials”, and OpenAI has confirmed nothing in what was read here.axios.com returns a 403 to this publication; the quotations come from a locally captured extract of the article carrying its canonical URL, author and timestamp, and that URL is cited.Axios itself states that it could not learn OpenAI’s expenses. That gap is the reporter’s own caveat, not this publication’s finding. Both growth figures are floors, so the true values may be higher. Nothing here is investment advice.
Axios reporter Madison Mills published a sourced scoop on September 29, 2026 at 14:15 UTC — canonical URL: axios.com/2026/09/29/… — reporting that OpenAI’s annualized revenue run rate has grown more than 70% since the beginning of the third quarter, reaching almost $70 billion, with enterprise sales more than doubling since July. This publication cites the canonical URL and read a locally captured extract of the article headed with its author and publication timestamp, because axios.com returns a 403 to this domain. OpenAI has confirmed none of these figures. The attribution rests entirely on sources described as familiar with the financials, and that is the whole of their provenance.
Both growth figures — more than 70% overall and more than doubled for enterprise — are floors. The true values may be higher; they cannot be lower. Working from those floors: a run rate nearing $70 billion after growth of more than 70% implies a starting run rate at the beginning of Q3 of roughly $41 billion or less. That is a ceiling derived by our own arithmetic, not a reported figure, and it compounds two approximations. It should be read as a bound, not a data point.
Axios placed its own caveat at the center of the story, and it is the most structurally important sentence in the piece: “Axios could not immediately learn details about OpenAI’s expenses, which is important context in relation to their businessgrowth.” That gap was identified and flagged by Axios — not discovered here. No expense, burn, margin, consumer-versus-enterprise revenue split, customer count, or IPO detail appears in what was read, and none is supplied in this article.
The key insight: A revenue run rate travels in a single sentence. The expense side requires a document. That asymmetry — not concealment — is why sourced scoops reliably surface growth figures and reliably do not surface margins. Axios named this gap itself. An IPO filing is the mechanism that forces both sides of the ledger into view at once, which is why Axios frames the moment around two companies preparing for one.
Indexed by this publication from the two growth figures Axios reports. Both are minimums, so the real positions sit at or above the orange dots.
The Structural Read
There is a well-worn pattern in how financial information leaks out of private companies: the top line escapes first, the cost structure escapes last, and the margin almost never escapes at all until a filing compels it. This is not a critique of anyone’s motives. It is a feature of information density. Someone familiar with revenue can repeat it in one sentence at lunch. Someone familiar with infrastructure spend, headcount cost, and compute depreciation would need a spreadsheet to do the same — and would be far more exposed in doing so.
The result is that every sourced revenue scoop on a pre-IPO AI company is structurally half an income statement. Axios said so plainly. That is worth holding onto: the number that circulates is the number that is easy to carry, not necessarily the number that is most decision-relevant.
Axios frames the broader moment this way: “The growth comes as OpenAI and Anthropic prepare for IPOs, which would give investors the clearest look yet at both the revenueopportunity and extraordinary spending driving their growth.” This publication has previously written up Anthropic’s prospectus figures as reported by Reuters — a case where a cost side existed precisely because a filing forced it out. Those figures are not restated here, and no comparison of the two companies’ economics is attempted, because this piece is not equipped to make one.
Axios — Sep 29, 2026
“Reality check: Axios could not immediately learn details about OpenAI’s expenses, which is important context in relation to their businessgrowth.”
Business Engineer Framework
The Map of AI — Where Does OpenAI Sit in the Stack?
OpenAI is the most visible application-layer company in the Map of AI’s nine-layer model. But application-layer revenue figures, however large, don’t describe the infrastructure layers underneath that consume the capital. Understanding which layer generates revenue and which layers consume it is the analytical move that sourced scoops cannot make for you.
Sources familiar with OpenAI’s financials have told Axios that the annualized revenue run rate is nearing $70 billion — a figure OpenAI has not confirmed, built from floors rather than fixed numbers, and missing the one variable that would make it analytically complete: what the company spends. Axios named that gap itself, and it stands as the most honest line in the story. Revenue at scale is a fact worth noting. Revenue at scale without a cost line is half a story, and on the evidence available today the other half is simply not in circulation.
Source: Axios — Madison Mills, Sep 29, 2026, 14:15 UTC (canonical URL cited; locally captured extract read, as axios.com returns a 403 to this publication). All figures are attributed to sources described as familiar with OpenAI’s financials. OpenAI has confirmed nothing reported here. The ~$41 billion start-of-Q3 ceiling is this publication’s own arithmetic and compounds two approximate figures. Nothing in this article is investment advice.
91,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
Every figure above is sourced reporting by Axios, attributed to people it describes only as familiar with OpenAI’s financials. It is not a company disclosure, and OpenAI has confirmed nothing in what was read here.axios.com returns a 403 to this publication. The quotations come from a locally captured extract of the article, carrying its canonical URL, its author Madison Mills and its publication timestamp of 29 September 2026 at 14:15 UTC. The canonical URL is cited above and the extract is what was read.The observation that the expense side is missing is Axios’s own reality check, quoted above and credited to Axios. It is not this publication’s finding. Both growth figures are floors rather than points — growth of more than 70 per cent, and enterprise sales more than doubled — so the real values may be higher and cannot be pinned. The figure of roughly $41 billion or less at the start of the third quarter is this publication’s arithmetic, is a ceiling rather than an estimate, and compounds two approximate figures. Nothing is established here about OpenAI’s expenses, burn, margin or path to profitability, about how revenue splits between consumer and enterprise, about any customer or seat count, or about the date, size or valuation of any public offering. Anthropic’s prospectus figures are not restated above and no comparison of the two companies’ economics is attempted. OpenAI’s DevDay keynote is scheduled for later today. Nothing above connects this report to it, describes it as timed, or previews any product. Nothing above predicts anything, and nothing here is investment advice.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
