The global race in artificial intelligence is typically framed as a contest over chips, talent, and models. Beijing is placing a major wager on a less glamorous input: electricity. The framing matters because training and running large AI models is energy-intensive work, which has turned access to cheap, reliable power into a strategic input alongside chips and talent.

According to a Bloomberg report, China aims to use its abundant clean energy to gain dominance in AI. The scale of the buildout is striking. From 2022 to 2025, China added more power-generation capacity than the entire US power system contains — not a comparison with a single year of American additions, but with the capacity of the whole US grid.

Moving the Computers to the Power

At the heart of the strategy is a government program called “East Data, West Computing.” The program responds to a basic geographic mismatch: demand for data sits in China’s crowded eastern cities, while the cheap electricity sits out west. Rather than hauling all that power east, Beijing wants to move the computing toward the power. The logic follows a simple cost reality: electricity ranks among the largest recurring operating expenses of a data center, so facility locations increasingly follow cheap generation.

China’s western regions are rich in solar, wind, and hydropower. Data centers placed there can, in theory, run AI workloads on cheaper and cleaner electricity. To organize the effort, the country has established eight national computing hubs and 10 data-center clusters — designated zones meant to spread computing resources across regions in a way that matches where electricity is plentiful and affordable.

Under the strategy, new data centers must source more than 80% of their electricity from clean sources. The plan also leans on expanding nuclear energy and building out ultra-high-voltage transmission lines, which are designed to carry electricity very long distances with less loss along the way.

The Numbers Behind the Appetite

Data centers currently account for about 1.68% of China’s electricity consumption, a share projected to reach 3% by 2030. The trajectory reflects the nature of AI infrastructure itself: clusters of power-dense accelerators running nearly continuously draw markedly more electricity than the conventional servers that host everyday web and cloud services.

Where the Plan Gets Complicated

China’s renewable expansion faces a known bottleneck: transmission infrastructure has not kept pace with generation. Clean electricity produced in a remote province does little good if the grid cannot move it, or if local demand to absorb it does not exist — a gap the industry describes as curtailment, when renewable output goes unused because it cannot be delivered or absorbed when and where it is produced. Bloomberg reports there are real questions about how effectively this power can be mobilized for AI.

Data centers need steady, dependable electricity around the clock. Solar and wind are not available around the clock. Hydropower can be steadier, but its output depends on the region and the season. Nuclear power and long-distance transmission are part of the answer, but both take time to build.

Distance cuts both ways as well. Placing computing in the far west lowers power costs, but it also puts servers further from the users and businesses in the east who rely on them. That trade-off maps onto how AI workloads differ: model training and batch processing tolerate distance far better than latency-sensitive services, which makes westward placement a more natural fit for some computing tasks than others.

What This Means

The 80% clean-electricity requirement ties AI growth to renewable growth by policy. If the mandate holds, every new data center becomes another customer for wind, solar, and hydro output in the west.

What to watch next: whether ultra-high-voltage transmission and nuclear capacity expand fast enough to stabilize supply, whether the eight hubs and 10 clusters attract meaningful AI activity, and whether data-center electricity use tracks toward the 3% projection by 2030.