China's AI models may be winning users by the million, but they are earning a tenth of what OpenAI and Anthropic make combined, according to estimates the Rhodium Group published on Thursday. The sharper point sits in the valuations: the research firm thinks investors are paying far more per dollar of revenue for DeepSeek and Moonshot than for either US leader.
Watch Desk analysis
What happened
Rhodium, a US-based research firm, compiled annual recurring revenue estimates for the major labs, using the industry habit of multiplying a recent month's revenue by twelve. The result is lopsided. All of China's AI models together generate about 10% of the revenue reported for OpenAI and Anthropic, whose figures Rhodium puts at $40bn and $65bn respectively.
On the Chinese side, DeepSeek sits lowest among the majors at $500m, followed by MiniMax at $800m and Moonshot at $1bn. Z.ai told investors on Wednesday that its latest ARR is $1.8bn and lifted its year-end forecast to $3bn. ByteDance's $4bn and Alibaba's $2.4bn lead the Chinese table and still trail the US leaders badly.
The multiples are where the report turns pointed. "Valuations relative to revenue appear exorbitant for Moonshot and DeepSeek at present," it says, citing estimated ratios of 50x and 163x against 34x for OpenAI and 21x for Anthropic. The timing is busy: Anthropic is reportedly expected to list in the US next month, OpenAI has pushed its IPO plans to next year, Moonshot has reportedly filed confidentially for a Hong Kong float and DeepSeek is reportedly preparing its own listing. Moonshot declined to comment on its filing, and DeepSeek and Anthropic did not respond to requests for comment, per CNBC's report.
The numbers
- $65bn: Anthropic
The largest figure in Rhodium's tally, ahead of OpenAI's $40bn. - $500m: DeepSeek
Lowest among major Chinese labs, on an estimated 163x valuation-to-revenue ratio. - $1bn: Moonshot
A 50x multiple, with a reported confidential Hong Kong IPO filing. - $1.8bn: Z.ai
Latest ARR, with the year-end forecast lifted to $3bn on Wednesday. - $4bn and $2.4bn: ByteDance and Alibaba
The largest Chinese figures, still a fraction of the US leaders.
Why it matters
Rhodium's Logan Wright argues the financing gap "means it will be far more difficult for Chinese frontier AI labs to scale sustainably", leaving them "heavily dependent upon a favorable climate in the equity market", which historically has not been an easy bet in China. He expects government funding to keep helping hardware and compute buildout while balking at direct subsidies for frontier labs. Rhodium also estimates that more than 60% of equity investment in Chinese AI chips and servers comes from state-affiliated sources.
There is a structural wrinkle underneath. Chinese labs mostly release open-weight models that anyone with capable hardware can run independently, and they are now exploring how to take a larger cut from third parties who serve those models. US models are mostly closed and cost far more per task, according to comparison firm Artificial Analysis. The labs giving the weights away are having genuine difficulty collecting rent on them.
The market mood is not helping either. US tech stocks have tumbled since American AI chiefs warned at the weekend about developing the technology too quickly, and leaders of the major Chinese labs have not yet commented. Z.ai shares still rose more than 5% in Thursday morning trading, recovering from this week's fall after news of its second big fundraise in two months.
Our read
Treat the precise figures as estimates, because Rhodium could only work from the latest available reports from this summer and ARR flatters anyone growing fast. The firm says as much, and Chinese usage has climbed steeply from a low base this year. None of that redeems a 163x multiple on its own.
The honest framing is that China's strongest open-weight labs have not yet built a business that converts adoption into revenue, and their valuations quietly assume that gets fixed. The good news for readers is that the story comes with its own test dates. Z.ai's $3bn promise now has a deadline, and Anthropic's reported listing would hand the market a US multiple to compare against. If you hold or covet any of these names, Rhodium has just published your diligence checklist.
What to watch
- Anthropic's reported listing
A US float next month would show what multiple public buyers pay against the 21x Rhodium cites. - The Hong Kong test
Moonshot's reported filing and DeepSeek's preparation will reveal whether markets accept the richer valuations. - Z.ai's $3bn promise
The year-end target was raised on Wednesday, so December offers a clean check on the forecast. - Open-weight monetisation
Whether Chinese labs find a way to bill for third-party access to models anyone can host.
Discussion spark: Rhodium calls DeepSeek's 163x multiple exorbitant, yet the open-weight labs are giving away what US rivals charge dearly for. Is Chinese AI under-monetised strategy or over-valued euphoria, and what number would change your mind?
Sources and evidence
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.