Chinese artificial intelligence (AI) models are spreading rapidly, but an analysis shows the pace at which they translate into revenue falls far short of U.S. AI companies. As Chinese AI firms' corporate values rise faster than actual profit scale, concerns about valuations are also growing.

On Sept. 17, CNBC reported that U.S. research firm Rhodium Group estimates annual recurring revenue (ARR) for all Chinese AI models at about 10 percent of the combined ARR of OpenAI and Anthropic. ARR is an indicator that estimates annual revenue by multiplying recent monthly revenue by 12. Rhodium Group put total ARR for Chinese AI models at about $10.7 billion, and cited OpenAI at $40 billion and Anthropic at $65 billion.

Among major Chinese AI companies, DeepSeek had the lowest ARR at $500 million, and Minimax was estimated at $800 million and Moonshot at $1 billion. Z.ai said on Sept. 16 that its latest ARR was $1.8 billion. Even adding ByteDance at $4 billion and Alibaba at $2.4 billion falls far short of OpenAI's $40 billion ARR on its own.

The problem is that valuations are rising faster than revenue. Rhodium Group assessed Moonshot and DeepSeek as currently looking excessive in terms of corporate value relative to revenue. The estimated multiples were 50 times for Moonshot and 163 times for DeepSeek, higher than OpenAI's 34 times and Anthropic's 21 times.

This gap also intersects with moves by major AI companies to list. Anthropic is pushing for a U.S. stock market listing, while OpenAI has postponed its initial public offering plan to next year. Moonshot is reported to have pursued a Hong Kong listing confidentially, and DeepSeek is also known to be preparing to list. Moonshot said it does not comment on market rumours or speculation regarding a confidential listing application.

Rhodium Group's analysis also has limits. The estimate is based on the latest figures secured up to this summer, and usage of Chinese AI models has surged recently from low levels early this year. Z.ai also raised its year-end ARR forecast to $3 billion from $2.4 billion.

Chinese AI companies are also exploring ways to monetise. Rhodium Group said Chinese AI labs are looking for ways to secure more revenue as third parties access their models to provide services. Chinese AI models are strongly open in nature, so with sufficient hardware they can be downloaded and run directly without going through developers. By contrast, most U.S. AI models are closed, and AI benchmarking firm Artificial Analysis says the main models of OpenAI and Anthropic also have higher per-task costs than Chinese models.

Logan Wright (로건 라이트), a partner at Rhodium Group, said it will be much more difficult for China's most advanced AI labs to continue scaling up because of a funding gap. Wright said Chinese AI companies have little choice but to depend heavily on a favourable stock market environment, but in China that has historically not been an easy option. He added that government funding helped build hardware such as expanding computing capacity, but it may be reluctant to directly fund cutting-edge AI labs.

Share price volatility was also high for listed Chinese AI companies. Z.ai shares rose more than 5 percent in morning trade on Sept. 17, recovering part of losses from earlier this week. Z.ai shares listed in Hong Kong rose more than threefold at one point in the summer but have recently fallen back to levels seen this spring. Rival Minimax has also struggled in recent months to hold on to first-day gains from its initial public offering.

Ultimately, the key challenge for China's AI industry has emerged as whether it can connect fast-growing model usage to actual revenue and sustainable growth.