China’s AI Boom Rewires Grid, Drives Data Centers to Deserts
China’s artificial intelligence boom is triggering a fundamental transformation of the country’s power grid, as surging energy demand from AI data centers forces a strategic shift toward desert-based computing hubs and the development of virtual power plants to manage electricity supply. The scale of the challenge — and the ambition of the response — is unprecedented.
The Datang Signal
When Chinese markets reopened after the May Day holiday in 2026, state-owned utility Datang International Power Generation Co. Ltd. embarked on a dizzying stock rally. After hitting the daily trading limit six times, its shares reached a record high of 9.92 yuan ($1.5) on June 3 — a surge of more than 130% in a single month, as Caixin Global reported.
The market frenzy was triggered by Datang’s launch of a 500-megawatt solar plant in Zhongwei, Ningxia — interpreted by investors as China’s first large-scale green-energy direct-supply project for computing. Yet the company issued eight risk advisories stating it had no operational “computing and power synergy” projects, underscoring the gap between market hype and operational reality.
The Scale of the Challenge
China’s computing electricity consumption has more than doubled from 82.4 billion kWh in 2019 to 196 billion kWh in 2025, rising from 1.3% to 1.9% of total power consumption. Projections from the China Academy of Information and Communications Technology indicate it could reach 500 to 700 billion kWh by 2030.
Driving this growth is an explosion in AI usage. China’s daily AI token usage surpassed 140 trillion by March 2026 — a more than 1,000-fold increase from early 2024, according to figures from the National Data Administration cited by CGTN.
The Core Tension: Stability, Not Volume
Wang Zesen, deputy director of the power system research institute at State Grid Jibei Electric Power Co. Ltd., told Caixin that “the core challenge of AI-driven computing growth is not total energy volume but instantaneous power delivery and grid stability.” AI inference workloads create a “peak upon a peak” effect, with servers causing instantaneous fluctuations up to 10,000 kW. Regions with high renewable generation and large data center loads face a “dual-high” stability risk — a problem illustrated dramatically in July 2024 when a thunderstorm in northern Virginia caused approximately 60 data centers to drop off the grid, losing 1,500 MW in seconds.
Gao Xing, chief utilities analyst at China Securities, offers a more measured perspective, arguing that “fears of AI causing widespread power shortages in China are overblown, as AI’s annual electricity demand increase is only about 1% of total load.”
Eastern Data, Western Computing
China’s strategic response is the “Eastern Data, Western Computing” initiative, launched in 2022. The program directs data centers to resource-rich western hubs — Ningxia, Inner Mongolia, Gansu, and Guizhou — where abundant renewable energy, cheaper land, and lower electricity costs offer compelling advantages. Data centers in these hubs must source over 80% of power from green energy.
Inner Mongolia exemplifies the strategy’s potential. As of May 2026, the region’s total computing power reached 315,000 petaflops — about one-seventh of the national total — including 297,000 petaflops of intelligent computing power, as Global Times reported. Electricity costs for computing centers are about 0.33 yuan/kWh, less than half the 0.76-0.80 yuan in eastern regions.
Li Chao, general manager of Huadian Inner Mongolia Energy, told Global Times that the first-phase 360-megawatt project is already in operation, delivering more than 700 million kWh of green electricity to data centers annually and reducing carbon dioxide emissions by 630,000 tons per year.
The $295 Billion National Computing Grid
China is drafting a five-year, 2 trillion yuan ($295 billion) program to connect thousands of data centers into a unified national computing grid, with a mandate that at least 80% of core technology — including AI accelerator chips — come from domestic suppliers, effectively locking out Nvidia and AMD, as TechTimes reported.
When power-grid integration is included, the total projected investment could reach at least 5 trillion yuan — approximately $740 billion. The grid is targeted for completion by 2028.
However, significant constraints loom. SMIC, China’s largest foundry, is limited to roughly 7nm process nodes due to export controls on EUV lithography equipment from ASML. SMIC’s co-CEO Zhao Haijun compared the rush to build data centers ahead of available chip supply to “building highways before the traffic arrives.” The Council on Foreign Relations projects the U.S.-China AI chip performance gap widening from 5x to 17x by 2027.
Virtual Power Plants and New Solutions
China is testing virtual power plants (VPPs) as a solution to grid flexibility challenges. In a landmark test, Shanghai migrated inference tasks to Fujian in three minutes to shed 50 kW of load. Guangdong integrated three telecom data centers into its electricity spot market via a VPP in May 2026.
Envision Group announced Mission Gobi at VivaTech 2026 on June 19 — a global initiative to build 5 GW of green AI data center capacity in desert and arid regions by 2030. Envision Chairman Zhang Lei stated that “even developing just 1% of the world’s desert and Gobi regions would be sufficient to support terawatt-scale computing capacity, at highly competitive costs,” as Sohu reported.
A Timeline Mismatch
A critical challenge remains the speed mismatch between infrastructure buildouts. Data centers can be constructed in 8 to 24 months, while substations take 3 to 5 years. Combined with opaque utilization rates ranging from 10% to 80%, grid planning becomes extraordinarily difficult.
Tian Feng, former dean of SenseTime’s Intelligence Industry Research Institute, told Global Times that the national computing network “is turning western green power into eastern computing capacity and helping convert China’s resource strengths into digital-economy competitiveness.”
What’s Next
On May 8, 2026, four top government bodies — including the National Development and Reform Commission and the National Energy Administration — issued a joint action plan targeting a secure energy guarantee system for AI by 2027 and full two-way empowerment by 2030. Both Alibaba and Tencent are also investing in next-generation nuclear technology, including small modular reactors and nuclear fusion startups.
Wang Yongzhen, an expert cited in the Caixin report, expects the experimental phase of computing and power coordination to last another three years. The ultimate goal: integrating security, green energy, and economic efficiency into a system that can power China’s AI ambitions without breaking the grid.
Whether China can resolve the tensions between chip supply constraints, grid stability risks, and explosive AI demand will determine not just the success of its digital transformation, but the global balance of AI competitiveness for the decade ahead.