Huge potential of tiny tokens powering AI revolution

In recent months, the term "token" has leapt from obscurity into everyday life as a buzzword alongside intelligent agents such as Lobster or Open-Claw. But this token is not the blockchain-based cryptocurrency unit.

Rather, it is the unit of measurement for large language model text processing. Just as we measure prose in letters or numbers, AI processes information by first breaking data down into bite-sized manageable tokens. Think of a token as the smallest Lego brick — the atomic particle — of AI-generated content.

Every time we use AI — whether asking a question or writing code — we invoke these atoms of AI language. Tokens, therefore, are an indicator of AI activity.

In this sense, token data not only reflect the dynamics of an economy in the AI era but also give rise to a new kind of economic logic.

At the beginning of 2024, China's average daily consumption of tokens was 100 billion. This shot up to a staggering 100 trillion by the end of 2025, and is now over 140 trillion.

This explosive growth of more than 1,000 times in a little over two years far exceeds the global average. Liu Liehong, head of the National Data Bureau, said, "The massive increase in daily token call volume indicates that China's AI development has entered a phase of rapid growth."

On the demand side, the token consumption frenzy indicates the widespread and frequent use of AI applications, reflecting the depth and breadth of AI empowerment across industries. It also points to a maturing ecosystem, hinting at a breakthrough in AI commercialization. As the world's largest internet economy, China is poised to be the epicenter of token economics.

On the other hand, this surge in token consumption reflects a supply-side momentum anchored in computing power. In the digital economy, data is the raw material while computing power is the production tool. Every token call consumes computing power.

The more powerful the computing capacity, the faster the token processing speed, the lower the unit cost, the higher the model quality, and consequently, the more frequent the token calls. The explosive growth in token consumption or call volume provides grounds for optimism regarding computing power supply prospects, while also sounding a warning that it is a scarce resource in the AI era.

With the development of the Metaverse, the Industrial Internet, and the proliferation of AI agents, the surge in demand might lead to a perpetual shortage in computing power. Computing power supply is constrained by investments in hardware and infrastructure such as chips, servers and data centers. With "AI+" becoming a national strategy, China must vigorously strengthen the supply-side infrastructure.

In the AI era, computing power is akin to national strength. Recent developments bode well for China's computing capacity. Beyond companies like Huawei Ascend, Cambricon, and Hygon Information Technology launching acceleration chips and CPUs, the completion of China's largest AI computing cluster in Zhengzhou is a milestone in domestic computing infrastructure.

Developed countries are racing to build AI for Science (AI4S) computing facilities, marking a new stage of global tech competition. The cluster in Zhengzhou, located at the core node of China's Supercomputing Internet, integrates 60,000 acceleration cards. It is a super brain tailored for Chinese scientists and signifies a breakthrough in the country's computing infrastructure.

What gives China greater confidence is the energy support behind computing power — the bottom layer of what NVIDIA CEO Jensen Huang calls the "Five-Layer Cake of AI". At the end of AI lies electricity; tokens are driven by watts.

China has the world's largest and most advanced power supply system, with a total installed generation capacity of 3.95 billion kilowatts — three times that of the United States and 13 times that of Japan. The continued expansion of green power promises the sustainable development of computing power.

Under the framework of token economics, China's two major initiatives — computing-electricity synergy project and "East Data, West Computing" — are increasingly significant. The former ensures power supply for computing centers through coordinated planning of computing infrastructure and power systems.

The latter channels dense computing power demand from the east to the electricity-rich west. Both are vital for consolidating and upgrading China's intelligent computing scale, now the second largest in the world.

As Liu of the National Data Bureau noted, "tokens are not only the value anchor of the AI era but also the settlement unit connecting technology supply with commercial demand, providing quantifiable possibilities for the implementation of business models."

Token economics is still in conceptual infancy, and its integration with the real economy has yet to crystallize into a mature model. Regardless of whether an industrial revolution centered on tokens has already begun, as some analysts speculate, establishing a new value assessment system around tokens — a core variable reshaping the digital economy landscape — is an inevitable requirement for the development of the AI industry.

The author is chief researcher at the China Watch Institute, China Daily.

The views don't necessarily reflect those of China Daily.

If you have a specific expertise, or would like to share your thought about our stories, then send us your writings at opinion@chinadaily.com.cn, and comment@chinadaily.com.cn.