Seven major Chinese AI model developers combined generate roughly $10.7 billion in annual recurring revenue, a fraction of the more than $100 billion combined ARR reported for OpenAI and Anthropic, according to estimates published by U.S.-based research firm Rhodium Group1,2.
The figures, drawn from reports covering March to August and using ARR as the metric, put the Chinese cohort at approximately 10% of the two leading U.S. AI companies.
Among Chinese developers, DeepSeek posted the lowest ARR at $500 million. MiniMax followed at $800 million, and Moonshot at $1 billion. Z.ai told investors its latest ARR stood at $1.8 billion, according to a transcript seen by CNBC. ByteDance's AI model ARR was pegged at $4 billion, and Alibaba's at $2.4 billion.
On the U.S. side, OpenAI's ARR reached $40 billion and Anthropic's $65 billion.
Rhodium flagged a disconnect between revenue and market valuations. The firm estimated Moonshot's valuation-to-revenue ratio at 50x and DeepSeek's at 163x. "Valuations relative to revenue appear exorbitant for Moonshot and DeepSeek at present," the Rhodium report said.
Z.ai raised its revenue forecasts on Wednesday, projecting its ARR will reach $3 billion by year-end, though that figure would still trail U.S. rivals by a wide margin.
Rhodium estimated that more than 60% of equity investment in Chinese AI chips and servers came from state-affiliated sources. ANALYSIS The heavy state role in infrastructure spending, paired with model-layer revenues that remain an order of magnitude below U.S. leaders, raises questions about the commercial sustainability of China's AI buildout independent of government capital.
The revenue gap is notable given the rapid adoption Chinese AI models have seen; high user growth has not yet converted into comparable monetization. DeepSeek's 163x valuation-to-revenue ratio, the highest Rhodium cited, underscores the extent to which investor pricing is running ahead of commercial traction.