Our country's daily average token call volume exceeds 140 trillion. Which investment opportunities are worth paying attention to?

Ask AI · How can China’s domestic computing power industry chain seize policy tailwinds?

China National Radio (CNR) Beijing, March 25 (reported by Li Shuo). According to China Media Group’s Economic Voice program “Trading in Real Time,” at a press conference held on March 24 by the State Council Information Office, Liu Liehong, director of the National Data Administration, revealed that by this March, the average daily call volume of Tokens in China has already exceeded 140 trillion. This figure is a more than 1,000-fold increase over the 100 billion at the beginning of 2024. Compared with the 100 trillion at the end of 2025, it has grown by another more than 40% within just three months.

Token, with its Chinese name of “ci yuan” (vocabulary token), is a core concept that has attracted significant attention in the field of artificial intelligence. This set of data intuitively demonstrates the rapid development momentum in this area. What exactly is Token? And what role does it play in everyday calls? Pan Lin, a member of the Information and Communications Economy Expert Committee of the Ministry of Industry and Information Technology, provided an illustrative analogy: Token is the “language atom” of the AI world.

He compares running AI to “eating.” Just as people need to eat dumplings one by one, when AI processes text and data, it will also first break the content down into one smallest unit that can be processed at a time—these units are Tokens. They can be a single Chinese character, an English word, or a punctuation mark, or they can be a fragment of a word.

Simply put, Token is the unit of measurement used to quantify the computing power and resource consumption during the process of using AI. The higher the Token call volume, the more frequently and at what larger scale users use AI. The “raising dragon shrimp” phenomenon that has recently drawn widespread attention is a typical manifestation of the rapid explosion of agent applications that is driving a surge in Token call volume.

In this regard, Wang Peng, deputy researcher at the Institute of Management Studies of the Beijing Academy of Social Sciences, said that in the early days, most Token calls were concentrated in the text generation domain. Nowadays, the call volume has grown significantly, mainly due to the concentrated surge of agent applications. Today’s artificial intelligence has formed a complete closed-loop capability from perception to decision-making to execution. Especially in real-world scenarios such as industrial manufacturing and warehousing/logistics, it has achieved high-frequency deployment and application. The “token economy” has entered a deep explosive period, and large numbers of business model innovations and technological innovations have been effectively validated.

At the press conference, Liu Liehong, director of the National Data Administration, also revealed that the next step for China will be to continue promoting data-enabled innovation in artificial intelligence development, and to coordinate with all parties to implement the new round of high-quality data set construction action plan in depth. Boosted by this news, on March 24, in the A-share market, the computing power leasing concept rallied in the afternoon. By the market close, multiple related concept stocks hit their daily trading limit.

Li Meixian, a senior analyst in the electronics industry at Hu an Securities, believes that sustained policy efforts, combined with the surge in Token call volume, will drive revenue growth across the computing power industry chain and will bring tangible benefits to China’s domestic computing power sector.

She said that as AI shifts from model training to large-scale inference applications, it will generate massive demand for computing power. In the future, computing power will be billed based on usage, just like electricity and tap water. At the same time, compared with overseas models, domestic models have clear competitive advantages in terms of power consumption and operations and maintenance costs, and this will translate into a definite demand for domestic computing power across the industry chain.

Zhuang Yang, a senior executive research specialist in the wealth management research department at Shenwan Hongyuan, also believes that starting in 2026, investors’ focus on the AI industry chain can gradually shift from overseas industry chains to domestic ones. In terms of investment operations, she suggests prioritizing issues whose price increases have been relatively limited, so that they can be built into a long-term position.

He said that from the perspective of the entire technology industry chain, growth certainty and continuity in the computing power leasing and computing power hardware sectors have strong support. Investors can focus on companies with a high degree of linkage to AI computing power but whose stock price has risen relatively modestly, and closely monitor the revenue structure of the relevant companies—especially those whose revenue from computing power business accounts for a high proportion—which have long-term holding value.