China's artificial intelligence models may be winning users at a remarkable pace, but new estimates suggest they are nowhere close to converting that popularity into revenue. According to CNBC, U.S.-based research firm Rhodium Group published estimates on Thursday showing that all of China's AI models combined generate only about 10 percent of the revenue reported by OpenAI and Anthropic.
The comparison is based on annual recurring revenue, or ARR, an industry metric that multiplies a recent monthly revenue figure by twelve to capture fast growth. On that basis, Rhodium estimates OpenAI alone at $40 billion and Anthropic at $65 billion. The figures underscore a widening commercial gap between the two countries' AI sectors, even as Chinese labs ship open-weight models that top global leaderboards. For more context on this story, see our ongoing latest AI developments.
The numbers behind the gap
Rhodium's estimates, as reported by CNBC, put the major Chinese labs' ARR in a range that starts startlingly low. DeepSeek, whose models have drawn international attention since early this year, posted the lowest figure among major Chinese AI companies at roughly $500 million. MiniMax followed at about $800 million, with Moonshot AI at around $1 billion.
Z.ai told investors on Wednesday that its latest ARR stood at $1.8 billion, according to a transcript seen by CNBC. Alibaba's figure was approximately $2.4 billion, and ByteDance led the Chinese pack at about $4 billion. Even combined, those totals trail the smallest of the two American leaders by a wide margin — and both OpenAI and Anthropic individually out-earn the entire Chinese field by roughly ten times.
Valuations look stretched in Beijing
The revenue shortfall would matter less if valuations had adjusted accordingly. They have not, Rhodium argues. The report describes valuation-to-revenue ratios for Chinese startups as "exorbitant" in the cases of Moonshot and DeepSeek, citing estimated multiples of 50x and 163x respectively.
By comparison, Rhodium estimates OpenAI at 34x and Anthropic at 21x. Those are still rich multiples by the standards of most industries, but they rest on revenue bases several orders of magnitude larger than their Chinese rivals'.
The funding environment compounds the problem. Logan Wright, a partner at Rhodium Group who co-authored the report with research analyst Endeavour Tian, said in a statement to CNBC that the financing gap "will make it far more difficult for Chinese frontier AI labs to scale sustainably." He noted that Chinese labs will be "heavily dependent upon a favorable climate in the equity market — historically that's not an easy bet in China," and predicted that government funding, while helpful for compute infrastructure, "will probably balk at direct funding for the frontier labs." Rhodium estimates that more than 60 percent of equity investment in Chinese AI chips and servers comes from state-affiliated sources.
IPO plans raise the stakes
The revenue revelations land at a delicate moment for both countries' AI champions. Anthropic is reportedly expected to list in the United States next month, while OpenAI has pushed its own IPO plans to next year. On the Chinese side, Moonshot has reportedly filed confidentially for a Hong Kong IPO, and DeepSeek is reportedly preparing for a listing as well. Moonshot declined to comment on market rumors, while DeepSeek and Anthropic did not respond to CNBC's request for comment.
Public markets have already given Chinese AI listings a rocky reception. Z.ai shares rose more than 5 percent in Thursday morning trading, recovering from declines earlier this week after news of its second major fundraise in two months. The Hong Kong-traded stock has fallen back to levels last seen this spring after briefly more than tripling over the summer. Rival MiniMax has struggled in recent months to hold above its IPO-day gains.
Z.ai separately told investors it now expects ARR to reach $3 billion by the end of the year, up from a previous forecast of $2.4 billion — a sign that Chinese labs believe the revenue picture is improving quickly.
Open weights, closed wallets
Part of the revenue gap is structural. Chinese labs have largely embraced open-source releases, which allow anyone with sufficiently capable hardware to download and run their models independently of the developer. That strategy spreads the models' influence but forfeits direct monetization. Rhodium noted that Chinese AI labs are exploring ways to capture more revenue from third parties offering access to their models.
American labs, by contrast, keep their leading models closed and charge premium prices. According to AI-comparison firm Artificial Analysis, the cost per task for leading models from OpenAI and Anthropic is far higher than that of their Chinese counterparts — pricing power that flows directly to the bottom line.
Rhodium cautioned that its analysis relies on the latest available figures from this summer, and that usage of Chinese AI models has skyrocketed from low levels earlier in the year. Even so, the firm's bottom line is blunt: China's AI sector has proven it can build frontier models, but it has yet to prove it can build frontier businesses.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
