Published : 17/07/2026
As the AI boom sweeps across the globe, this surge in demand has naturally pushed up the prices of AI models and tokens.
Will the cost of using AI continue to rise? Will AI remain affordable for everyone in the future? China is actively promoting the construction of a national "computing power network", with the ultimate goal of making computing power as universally accessible and affordable as water, electricity, and the internet.
Computing power network included in China's top-level strategic planning
What is the demand for tokens in China? Let us first look at a set of figures. In March 2026, China's average daily token usage exceeded 140 trillion, an increase of more than 1,000 times compared to the beginning of 2024.
With demand outstripping supply, prices have naturally soared. Recently, the prices of some Tencent Cloud AI models have increased by over 400%, while Alibaba Cloud adjusted its prices three times in a single month between March and April.
Will the price of tokens rise endlessly? Will we still be able to afford AI in the future? Many people have this concern.
In response, China has proposed to build the computing power network.
The 15th Five-Year Plan proposes to build new types of infrastructure proactively and ahead of schedule, which includes constructing a multi-level computing power facility system and a national integrated computing power network.
At recent meetings of the Political Bureau of the Central Committee and executive meetings of the State Council, frequent mention has been made of the "six networks", namely to strengthen the planning and construction of the water network, new-type power grid, computing power network, new-generation communications network, urban underground pipe network, and logistics network.
It is evident that the computing power network is not only included in China's top-level strategic planning but is also being compared with public infrastructure like water and electricity, placing it in a position of equal importance.
What is the computing power network? And why build it?
So, what exactly is the computing power network? Simply put, it is a "computing power version of the national power grid".
When we use electricity, we do not need every household to install its own generator because the main power grid handles unified construction and distribution.
The computing power network is the same. It connects data centres and supercomputing centres across the country into a single network for unified scheduling.
The current use of computing power can be likened to each household building its own power station and using its own electricity; whereas the computing power network is like building the power grid, allowing electricity to flow through standard interfaces and be scheduled uniformly.
The difference is not just about connecting more machines, but about turning computing power into a new type of measurable and tradable infrastructure.
The reason for building a nationally unified computing power network is mainly because the spatio-temporal distribution of computing power resources is currently unbalanced.
The computing power load in the east is saturated, and supply is critical; whereas in the west, high-quality green electricity computing power resources are idle and wasted.
Inefficient cross-regional scheduling and prominent resource mismatch issues mean that the traditional computing power model can no longer adapt to the rapid pace of development of the AI industry.
An expert has pointed out that, "China has a vast number and scale of various types of computing power centres, yet these centres are mostly scattered and operate autonomously, forming numerous 'computing power islands'. Building a nationally unified computing power network can change this situation."
What changes will the computing power network bring?
So, specifically, what kind of changes will it bring after the computing power network is built?
To summarise in one sentence, it is "to make computing power easier to use, cheaper, and more equitable", and this has practical dividends for individuals, companies, and the country.
For individuals, the goal of the computing power network is to allow individual users to use computing power as if they were using mobile data. Any access point can be matched with suitable computing power through an intelligent system. When using AI tools, the responses will be faster and more stable, and waiting times and lag will be reduced.
For companies, the computing power network allows them to rent computing power from the entire network on demand, instead of investing large amounts of capital to build their own computer rooms and servers. This allows companies to focus more on their business and models, rather than on hardware and maintenance.
At the national level, it can strengthen global competitiveness. The scale of computing power is becoming a vital metric for measuring a country's digital economy and technological competitiveness.
A unified national computing power network helps consolidate resources and build a national-level advantage in computing power, providing support for long-term competition across multiple fields, including large language models (LLMs), supercomputing applications, and national defence research.
Read more:
What is a "Token factory"? What advantages does China have?
How can China's token exporting be realised?
What is computing-electricity synergy and why is it important to AI?