China’s AI Revenue Lags Behind U.S. Competitors, Raising Valuation Concerns
- China’s AI models generate only 10% of the revenue compared to OpenAI and Anthropic, according to Rhodium Group.
- Valuations for Chinese AI firms like DeepSeek and Moonshot are significantly higher than those of their U.S. counterparts.
- Z.ai forecasts its ARR to reach $3 billion by year-end, up from a previous estimate of $2.4 billion.
BEIJING — While China’s artificial intelligence models have seen swift adoption, this increase has not yet translated into substantial revenue, stirring skepticism about the valuations of local companies. According to estimates released on Thursday by U.S.-based research firm Rhodium Group, the combined revenue from all of China’s AI models is merely 10% of that reported by leading firms OpenAI and Anthropic.
Using the industry metric of annual recurring revenues (ARR), which assesses growth by annualizing a recent monthly figure, the report outlines that DeepSeek’s ARR stands at $500 million, the lowest among major Chinese AI companies. MiniMax follows with an ARR of $800 million, and Moonshot reports $1 billion. Z.ai provided an update to investors on Wednesday, revealing its latest ARR at $1.8 billion.
Despite notable figures for ByteDance at $4 billion and Alibaba at $2.4 billion, all these revenues pale in comparison to the staggering $40 billion generated by OpenAI alone, not to mention Anthropic’s $65 billion. This low revenue generation raises concerns about how investors are currently valuing Chinese startups.
The Rhodium report pointed out that the revenue-to-valuation ratios for companies like Moonshot and DeepSeek appear inflated, at 50x and 163x, respectively, compared to OpenAI’s 34x and Anthropic’s 21x ratios.
Z.ai’s Revenue Forecast
Importantly, the Rhodium analysis is based on the most recent available data from this summer, while the use of Chinese AI models has surged from earlier in the year. Z.ai has announced an updated forecast, expecting its ARR to reach $3 billion by year-end, a revision from an earlier estimate of $2.4 billion.
Rhodium has also noted that Chinese AI labs are investigating methods to secure a larger share of revenue from third-party access, as the open-source nature of their models enables anyone with sufficient hardware to operate them independently. In contrast, U.S. AI models are predominantly closed-source, with cost analyses showing that the cost per task for leading models from OpenAI and Anthropic far exceeds that of Chinese offerings.
Logan Wright, a partner at Rhodium Group and co-author of the report, highlighted the “financing gap,” suggesting it complicates the sustainable scaling of Chinese frontier AI labs. He noted the companies’ reliance on favorable equity market conditions, which have historically been challenging in China. Government funding has aided in hardware development, but may hesitate to provide direct support to frontier AI labs.
Rhodium estimates that over 60% of equity investment within China’s AI chips and servers originates from state-affiliated sources.
This year has been tumultuous for Chinese AI firms that have gone public. Z.ai’s shares rebounded by more than 5% in morning trading on Thursday after declining earlier in the week, following its second significant fundraising round in two months. However, the stock has dropped back to levels reminiscent of this past spring after experiencing over a threefold increase during the summer.
Minimax, a competitor, has faced difficulties in maintaining its initial public offering-day gains after a period of price spikes in the spring.
News Courtesy of CNBC
