The 2026 Government Work Report included "computing-power and electricity synergy" in the new infrastructure projects for the first time, and the "15th Five-Year Plan" also explicitly proposed "promoting the coordinated layout of green power and computing power." Liu Liehong, Director of the National Data Administration, pointed out at the China Development Forum held in March this year that the essence of computing-power and electricity synergy is "deeply integrating computing power infrastructure with the power system through digital technologies, intelligent algorithms, and information networks, promoting dynamic matching and optimized allocation of resources."
As AI (Artificial Intelligence) demand for computing power climbs exponentially, the power consumption of computing power is also surging simultaneously. Data from the National Energy Administration shows that in 2025, the power consumption of China's computing power facilities reached 170 billion kW·h, a year-on-year increase of about 30%. It is expected to exceed 700 billion kW·h by 2030, accounting for 6% of the total power consumption of the whole society. The relationship between computing power and electricity is shifting from "one-way power supply" to "two-way synergy."
Under the proposition of this era, Qingyang, Gansu—a prefecture-level city in the northwest with a permanent population of only 2.1 million—is writing a unique "Qingyang Solution" for computing-power and electricity synergy.
Physical Integration: Green Power "Direct Supply via Dedicated Lines" Restructures Computing Power Costs
The computing power industry is a "power-hungry industry". Public data shows that a single training session of GPT-4 consumes as much as 240 million kW·h of electricity. As AI large models advance from tens of billions of parameters to trillions of parameters, the proportion of electricity costs in computing power operating costs has climbed from 30% to 50%–70%. For Qingyang, this exploration of computing-power and electricity synergy begins with changing the power supply mode.
Qingyang's developable scale of wind and solar new energy exceeds 22.2 million kW. Coupled with natural endowments such as low annual average temperature and stable geological structure, it possesses inherent advantages for developing green computing power. However, traditional new energy power generation is highly volatile, making it difficult to directly match the stable 7×24-hour power demand of data centers. The change began in November 2024, when Gansu Electric Power Investment Group was approved for the country's first pilot project for aggregated direct supply of green power dedicated to data centers. It aggregates scattered wind and solar new energy in the area, equipped with energy storage and synchronous condensers to smooth out fluctuations, realizing the direct delivery of stable green power to the China Computing Valley Industrial Park.
"Previously, wind power and photovoltaics were connected to the grid independently, with highly fluctuating output. It was impossible to match the 7×24-hour stable power supply demand of data centers, leaving a large amount of clean power idle and wasted. Now, we aggregate scattered new energy in the area, equipped with energy storage and synchronous condensers to smooth out fluctuations, and deliver stable green power directly to the China Computing Valley Industrial Park via dedicated lines," introduced a relevant person in charge of Gansu Electric Power Investment Group. 55% of the park's electricity is directly supplied by local wind and solar power, and the remaining power is made up through power market transactions. From January to May 2026, a total of over 13 million yuan was saved in electricity costs for resident enterprises, and the cost reduction scale for the whole year is expected to exceed 30 million yuan.
Chen Changliu, General Manager of the Computing Power Branch of China Mobile Gansu Company, calculated an account: the park is equipped with a 2 million kW wind and solar green power base and energy storage facilities. Green power is directly sent to the data center through a self-built 110 kV substation. The proportion of green power in the park exceeds 80%, and the end-user electricity price is stabilized at 0.398 yuan/kW·h—about half of the industrial electricity price in the eastern region, and 0.2 to 0.3 yuan lower than the average electricity price of data centers in the east. "10,000 PFLOPS of computing power can save tens of millions of yuan in electricity bills a year, which is a tangible attraction for computing power enterprises."
Lei Qiang, Director of the Qingyang Project Department of Gansu Electric Power Investment Jiuhui Company, pointed out that the total scale of the project is 2 million kW. The core model is the deep synergy of "computing power + electricity". Relying on Qingyang's local wind and solar resources, an integrated system of "source-grid-load-storage" is built to realize nearby development, aggregated trading, local consumption, and precise direct supply of new energy. Different from the traditional unified power supply mode of the power grid, this project prioritizes matching the computing power load of the park. 55% of the park's electricity consumption achieves direct supply matching of generation and consumption within the site, and the remaining 45% is made up through medium- and long-term power transactions and the spot market, maximizing the reduction of the comprehensive electricity price of the park. Since the first phase of the project was put into operation, it can save 616,700 tons of standard coal and reduce carbon dioxide emissions by 1.6157 million tons annually, perfectly meeting the low-carbon transformation requirements of data centers.
Dispatch Synergy: Making the Most of Every Kilowatt-Hour of Green Power
The "direct supply" at the physical level is only the first step; the deeper synergy lies at the dispatch level—how to make computing power proactively "adapt" to the rhythm of green power.
More cutting-edge explorations are underway. China Mobile Gansu Company has built a triple power supply mode in the park: "green power direct supply as the bottom line, market-oriented power as the supplement, and large power grid security guarantee." When wind and solar output is sufficient, priority is given to direct supply via dedicated lines; when power generation exceeds the load, surplus power is fed into the grid with shared benefits; during fluctuating periods, procurement is made through the power spot market to fill the gap. These three guarantees ensure that computing power enterprises use "reassuring power."
China Mobile Gansu Company and Gansu Electric Power Investment are continuously deepening computing-power and electricity synergy. On the one hand, based on the existing grid-connected power supply, they are exploring the "source-grid-load-storage" off-grid supply mode, accelerating the construction of a 330 kV dedicated line to directly connect the wind and solar green power in Huanxian, Huachi, and other places to the park, gradually improving the off-grid operation capability of computing power clusters, and striving to achieve "full green power supply for computing power," with the expectation that the end-user electricity price will reach 0.36 yuan/kW·h. On the other hand, they are jointly tackling the "computing follows power dispatch" mode with green power suppliers and computing power demand enterprises—accurately dispatching computing power tasks according to the peak and valley of green power output. When green power is abundant, training tasks are fully "consumed"; when green power is tight, the rhythm of non-urgent tasks is flexibly adjusted. The goal is to make computing power follow green power and maximize the benefits of every kilowatt-hour of green power.
The dual efforts of off-grid operation and "computing follows power dispatch," if promoted on a large scale, will have profound significance for new energy consumption nationwide. "Looking to the future, China Mobile Gansu Company is actively exploring the 'computing-power-electricity-carbon' trading mode, attempting to integrate computing power dispatch, green power consumption, and carbon asset operations into the same set of market mechanisms, so that the value of carbon reduction can be measured, traded, and amplified," said Chen Changliu.
Industrial Ecology: Low Electricity Price Dividends Spawn Agglomeration Effects
The dividends of computing-power and electricity synergy are ultimately reflected in the "voting with their feet" of the industrial ecology. Since Qingyang was identified as one of the eight national hub nodes for the "Eastern Data and Western Computing" project in 2021, China Mobile Gansu Company took the lead in responding, investing over 2.4 billion yuan in total, building 4 data centers with a bearing capacity of 92,000 PFLOPS of computing power, and achieving a cabinet mounting rate of 85%. As of now, the total computing power scale of the Qingyang data center cluster has reached 185,500 PFLOPS, with intelligent computing accounting for over 99%, representing 97% of the province's computing power and 9% of the nation's. Qingyang has become a benchmark among the eight national hub nodes with the fastest computing power growth rate, the largest increment, and the highest proportion of intelligent computing.
Low electricity prices are the strongest "investment invitation letter." Relying on stable green power supply and significant cost advantages, China Mobile has introduced more than 70 industry-leading partners, including Kingsoft Cloud, Alibaba, MiniMax, Zhipu, WPS, Enflame Technology, and Meitu. Through the mode of attracting business through business, a total of over 10 billion yuan of investment has been introduced in Qingyang. Large models such as DeepSeek, MiniMax, and Zhipu Qingyan, which rank among the top six globally in terms of call volume, are all deployed in Qingyang, bearing more than 25% of the independent model Token computing demand nationwide.
In terms of transmission guarantee, China Mobile Gansu Company has built Gansu's first 400 Gb/s backbone OTN. The latency from Lanzhou to Qingyang has been reduced to within 3 ms, and the inter-provincial latency to Beijing-Tianjin-Hebei is 8 ms, to the Yangtze River Delta is 12 ms, and to the Guangdong-Hong Kong-Macao Greater Bay Area is 15 ms. Qingyang has achieved "one-hop direct access" to key cities across the country. Data published by the Ministry of Industry and Information Technology in 2024 shows that the transmission performance from Qingyang to the other seven hub nodes is the best in the country, smoothing the "main artery" of computing power transmission for "Eastern Data and Western Computing."
Model Enlightenment: Computing-Power and Electricity Synergy from "Qingyang Practice" to "National Solution"
Qingyang's exploration has provided replicable experience for computing-power and electricity synergy nationwide. From the physical layer perspective, the mode of "nearby development of new energy, aggregated trading, local consumption, and precise direct supply" has effectively resolved the contradiction between the volatility of new energy and the stability of data centers. The core of this mode lies in "aggregation"—coordinating the development and centralized dispatch of scattered new energy projects to form a stable and controllable green power supply capacity. Compared with the mode of purchasing green power certificates or scattered procurement of green power, direct supply via dedicated lines has significant advantages in both cost and stability.
From the dispatch layer perspective, the exploration of "computing follows power dispatch" and "source-grid-load-storage" has opened a new path for the two-way synergy of computing power and electricity. When computing power load becomes an adjustable power demand resource, new energy consumption will gain a "super stabilizer." The China Electricity Council predicts that by 2030, the adjustable load potential of national computing power clusters will exceed 20 million kW, equivalent to the regulation capacity of 10 million-kW-class thermal power units.
From the value layer perspective, low-priced green power reduces computing power costs, while the computing power industry provides scenarios and returns for new energy consumption, forming a virtuous cycle of "electricity supporting computing, and computing feeding back electricity." Qingyang's practice proves that as long as the mode is reasonably designed, the green power advantages in the western region can be fully transformed into competitive advantages in the computing power industry.
On the loess tableland, wind turbines are turning, photovoltaics are spreading, and servers are humming—a picture of the "two-way pursuit" of green power and computing power is gradually unfolding in the Longdong land. This is not only Qingyang's exploration but also a vivid footnote to China's computing-power and electricity synergy moving from "concept" to "paradigm."
Authors: Tian Yongbin, Shi Zheming, Sun Fadun, China Mobile Gansu Company This article is published in the 8th issue of "Telecom World" in 2026 -END- Managing Editor/Layout: Zhu Wenfeng Proofreader: Mei Yaxin Supervisor: Liu Qi