At the 2026 World Artificial Intelligence Conference, the National Energy Administration issued a clear signal: China will accelerate the construction of a new energy system to provide a green foundation for the development of artificial intelligence. Green electricity accounts for nearly 40% of every kilowatt hour of electricity, while computing power consumption is skyrocketing at an average annual rate of nearly 40%. When the usage of tokens skyrockets by a thousand times in two years, how to make AI computing power eat cheap and clean green rice has become a strategic must answer question from a technical multiple-choice question.

From 170 billion to 800 billion degrees in five years

The computing power industry has become the fastest engine driving the growth of electricity consumption in the whole society. According to data from the National Energy Administration, the electricity consumption of computing power facilities in China will reach 170 billion kilowatt hours by 2025, a year-on-year increase of about 30%, accounting for about 1.6% of the total electricity consumption in the society. The annual average growth rate of electricity consumption in the eight major nodes of national integrated computing power has reached 39.5% in the past three years, far exceeding the average growth rate of electricity consumption in the whole society. The China Electricity Council predicts that by 2030, the national computing power consumption will reach 800 billion kilowatt hours, accounting for about 6% of the total electricity consumption in society, with an average annual increase of over 100 billion kilowatt hours during the 15th Five Year Plan period.

At the same time, the support capacity of the green power base is synchronously improved. As of the end of June 2026, China's cumulative installed power generation capacity has exceeded 4 billion kilowatts, and the installed renewable energy capacity has exceeded 2.4 billion kilowatts, continuing to rank first in the world. Wang Hongzhi, Director of the National Energy Administration, stated at the forum that nearly four out of every ten kilowatt hours of electricity consumption in the whole society is green electricity. Let every kilowatt hour of electricity be fully utilized and every token go hand in hand with green energy.

Computer Computer Collaborative Acceleration Run

During the 15th Five Year Plan period, AI inference load will exceed training load and become the main consumption of computing power. New features such as time delay sensitivity, large load fluctuations, and concentrated demand require a shift in computing power collaboration from spatial collaboration to multidimensional collaboration between computing power loads and new power systems. In 2026, the collaboration of computing and electronics will be included in the government work report for the first time, and the four departments of the country will subsequently jointly issue an action plan to deploy ten tasks including computing and electronics collaboration.

From the green power direct supply data center in Helingeer, Inner Mongolia, to the cross provincial computing power output from Yiwu, Xinjiang to Chongqing, and to the integrated model of source network load storage that was first implemented in Ulanqab, computing power collaboration is moving from policy documents to construction sites on the Gobi Desert. This is a deep transformation driven by national strategy, AI demand, and green power resources.Editor/Cheng Liting