As U.S. and Chinese AI companies rush to roll out new models, interest is growing among the tech industry and users in how much AI performance can improve. Against that backdrop, some observers say price competition will also be a key point to watch in the race between U.S. and Chinese AI firms. That view rests on the idea that Chinese tech companies have narrowed the gap significantly with U.S. rivals. If performance is broadly similar, price could decide the outcome.
On price, Chinese tech companies may have an edge. A recent report by the Financial Times (FT) said the entry-level plan for an AI chatbot offered by Chinese AI startup ZhipuAI (Z.ai) costs about $3 a month. That is a steep discount compared with OpenAI's ChatGPT and Anthropic's Claude, which are close to $20 a month. Using the phrase “price war,” the FT said that if performance is similar, AI prices are likely to be determined before long by the cheapest model that is “good enough to use.”
At the current point, U.S. AI companies have higher valuations than Chinese firms. Since ZhipuAI debuted on the Hong Kong stock market, its share price has jumped 300 percent and its valuation has risen to $29 billion, but it still falls short of U.S. companies. U.S. firms have a much larger revenue base than Chinese rivals and remain better positioned in ties with established companies and access to capital.
But the AI technology gap between U.S. and Chinese companies is narrowing. On standard industry benchmarks, major Chinese LLM companies are operating at levels similar to competing U.S. models across reasoning, coding and general knowledge tasks, the FT reported.
According to the FT, ZhipuAI has trained its AI models on Huawei Ascend AI chips, which lag Nvidia’s latest GPUs, after the United States restricted exports of AI chips to China. But the export restrictions have not eliminated China’s ability to build reliable AI models. That is especially the case if computing chip design and software improvements make up for lower computing performance.
Beyond ZhipuAI, Chinese tech company Alibaba recently unveiled a new AI model, Qwen 3.5. Qwen 3.5 is designed to carry out complex tasks independently and has outperformed major U.S. rival models on multiple benchmarks in terms of performance and cost, the company said. ByteDance released an AI-based video generation model, Seedance 2.0. Seedance 2.0 is drawing reactions from users calling it a “game changer” for features such as multi-scene processing, audio and dialogue synchronisation, and multilingual support.
If Chinese companies are offering AI models comparable to those from U.S. firms, the weight of price can only grow.
OpenAI charges developers $1.75 per 1 million input tokens to access its flagship model GPT-5.2 and $14 for output tokens. Tokens are the text units processed by AI models, and typically about 4 characters are considered 1 token. By contrast, ZhipuAI's GLM-5 costs $0.58 for input tokens and $2.60 for output tokens. The gap is pronounced on output.
The FT highlighted ZhipuAI’s price competitiveness and said user behaviour could change in the future even in markets where U.S. companies have an advantage in brand awareness and distribution. It said the impact could be larger for students, independent developers, small and medium-sized businesses and cost-sensitive startups.
According to market research firm Counterpoint Research, consumer spending on generative AI is forecast to reach $700 billion by 20230. The figure excludes enterprise and cloud infrastructure revenue.
If Chinese AI companies secure one-third of the market and ZhipuAI takes 10 percent of that, the company could reach $23 billion by 2030.
ZhipuAI also appears to be seeking to expand its territory in the global market by leading with price. Its strategy is to compete with high-performance, low-cost models. A January report by Bloomberg said ZhipuAI forecast that fierce price competition, after it cut API costs in China by about 85 percent, would spread overseas as well. Bloomberg said this would challenge U.S. rivals such as OpenAI.