OpenAI's Annual Recurring Revenue Approaches $70 Billion
nashnova research
OpenAI's ARR is nearing $70 billion, up over 70% since the start of Q3 with enterprise revenue doubling; but spending data remains undisclosed, leaving the true cost of that growth unknown.
$70 billion in ARR — how fast is that?
Sources told Axios that OpenAI's annual recurring revenue (ARR — the full-year run rate based on current paying customers) is approaching $70 billion, up more than 70% since the start of Q3.
Enterprise (B2B) revenue more than doubled over the same period. This means → big corporate clients are adopting faster than consumers, and OpenAI is shifting from "chat tool" to "enterprise infrastructure."
Consumer revenue is surging too: new consumer revenue added in Q3 alone already exceeds all of 2025's consumer revenue.
In plain terms = growth isn't riding a single engine — enterprise and consumer are accelerating in parallel, a "dual-engine" commercialization phase.
Growth this fast — why can't we call it yet?
Axios explicitly noted: OpenAI's spending details are not available.
This means → we're only seeing the "money in" half; "how much it cost to earn that money" is a complete black box.
For AI companies, compute, training, and infrastructure burn rates often match or exceed revenue growth — without spending data, the revenue number is only half the story.
Can Anthropic keep up?
Anthropic long dominated enterprise AI adoption, but OpenAI's doubling of enterprise revenue signals the gap is closing fast.
Per Reuters, citing Anthropic's IPO prospectus, its 2025 revenue grew 12× to nearly $4.6 billion; per the New York Times and others, its annualized revenue reached roughly $65 billion by July and continues to climb.
Yet Anthropic's prospectus also disclosed future cloud, compute, and infrastructure contractual obligations totaling $518 billion. In plain terms = revenue is soaring, but locked-in future spending is an even larger number.
Anthropic further warned investors in the prospectus that its technology could pose an "existential risk" to humanity — an extraordinarily rare statement in an IPO filing.
Does the real exam start after the IPO?
Both OpenAI and Anthropic are preparing for IPOs that will force full disclosure of revenue and cost structures for the first time.
This means → the market is currently working with "selective data" — revenue figures released voluntarily by the companies, while spending figures sit in undisclosed books.
This reflects a deeper issue: current AI-sector valuations rest on only half the information. The IPO is the real open-book exam — and when it arrives, the cost pressures behind headline growth will have nowhere to hide.
市场有风险，内容仅供研究参考，不构成投资建议。