In China's artificial intelligence (AI) industry, a token-based revenue model is spreading rapidly. Analysts say the so-called "token economy" has entered a full-fledged growth phase, especially as token usage jumped more than 40% in the past three months.
A token is the smallest unit used when an AI model processes data. Generally, one token corresponds to one English word. During AI service use, inputs and outputs are split and calculated in this unit, and major corporations charge fees based on token usage.
According to China's National Data Administration on the 25th, as of March, China's daily token call volume exceeded 140 trillion. That is an increase of more than 40% in just three months from 100 trillion at the end of 2025. Compared with the beginning of 2024, when it was 100 billion, it has increased more than 1,000-fold in two years. Some AI model corporations saw their revenue over 20 days last month surpass their annual revenue for 2025. As the industrial importance of tokens rose, the Chinese government on the 24th proposed "cıyuán" as the official translation for token.
The Chinese government defines tokens not as a simple technical concept but as an economic unit. According to the Chinese business outlet Cailianshe, National Data Administration Director General Liu Liehong said, "Tokens are a settlement unit that links technological supply and commercial demand," adding, "A new value system centered on the calling, distribution, and settlement of tokens is rapidly taking shape." He went on, "Tokens have attributes that can be measured, priced, and transacted," and "they could become a key path for the AI industry to lead to actual revenue." In practice, the explanation is that the more token usage increases, the more service use grows, which in turn leads directly to revenue expansion.
The rise of this "token economy" is being felt in industry as well. Alibaba recently officially launched the Alibaba Token Hub business unit. The organization's core goals are the production, distribution, and utilization of tokens. Chief Executive Officer (CEO) Wu Yongming is personally leading it. It is an attempt to shift the AI business from model-centered to token-centered. At major corporations such as U.S. Meta and OpenAI, there are cases in which token usage per employee is managed and a "token budget" is provided to employees as a benefit to advance and automate work. Cailianshe assessed, "A virtuous cycle in which data supply leads to value creation has begun to take shape."
According to China Business News, the Chinese government is also supporting this with policy. As the token economy quickly takes hold, it is accelerating efforts to increase the high-quality data that underpins the AI industry. The gist is to ▲ expand the development and distribution of data resources, ▲ build a unified national data market, and ▲ establish a property rights system that separates data ownership, usage rights, and operating rights. Efforts to expand the supply of high-quality data are also gaining steam. As of the end of 2025, the number of high-quality datasets built by China exceeds 100,000, and their scale is about 310 times that of the National Library of China's digital resources. The National Data Administration plans to expand the high-quality dataset framework with 26 ministries and increase the supply of datasets that can be used industrially.