Chinese AI models grow in use, but many makers struggle to profit

In short: Chinese AI chat models are being adopted quickly in China and overseas because they work well and cost less, but many companies behind them still do not have a clear way to make steady profits.

What's going on

Chinese large language models, which are the text engines behind many chatbots (like a writing and reading assistant), are spreading fast. Models like DeepSeek, Alibaba’s Qwen, and Moonshot’s Kimi are often described as close in quality to top US models, but cheaper to use.

Price is a big reason. Some estimates say Chinese models can cost 60% to 90% less than leading US closed models, and one benchmarking group says the cost can be about one-fifth for many users.

Another driver is that many Chinese models are “open-weight,” meaning people can download the model and run it on their own computers or servers (like buying the recipe, not just ordering the meal). The licenses are often flexible, so companies can modify the model and use it in paid products.

Chinese models are also showing up in products outside China. For example, the US-based coding tool Cursor adopted a Kimi model as a core engine, which points to overseas licensing deals as one possible source of revenue.

What to watch

The big challenge is money. When a model is freely downloadable, the original creator may not see how it is used and cannot easily charge per use or collect feedback to improve it. Companies are trying other paths, like selling cloud hosting, charging for usage through paid access, offering subscriptions in consumer apps, and selling business-focused tools. The question is which of these will turn widespread use into reliable profit.

Source: NYTimes