China’s Dominance in Power Generation Underpins Its Edge in the AI Race

By John Ian Lau, Contributing Editor

BEIJING—As the global race for artificial intelligence supremacy intensifies, one critical factor is emerging as the decisive battleground: electricity. Training and running advanced AI models demand enormous computational resources, which in turn require vast amounts of reliable power. China, with its unparalleled expansion of electricity generation capacity, is pulling ahead in this underlying energy contest.

China’s installed power generation capacity reached a record 3.75 terawatts as of late 2025, according to industry analyses and official projections trending toward 3.9 terawatts by year-end. This marks a near-doubling over the past eight years, driven by aggressive investments in renewables, coal, and nuclear. The scale is staggering: China’s capacity is nearly three times that of the United States, which stands at approximately 1.3 terawatts.

This energy abundance provides China with a structural advantage in AI development. Hyperscale data centers powering large-language models can consume gigawatts of electricity—equivalent to the needs of entire cities. Abundant, low-cost power allows Chinese firms to scale AI training and inference at a pace that constrained grids elsewhere cannot match. Analysts have dubbed this the “electron gap,” where China’s oversupply of generation capacity enables brute-force computing strategies that remain economically prohibitive in power-constrained markets.

The disparity extends to nuclear power, a baseload source ideal for the constant demand of AI infrastructure. China has around 30 reactors under construction, accounting for nearly half of global nuclear builds. Hundreds more are planned or proposed. In contrast, the U.S. has no large-scale commercial nuclear reactors currently under construction, following years of delays and cost overruns on recent projects.

This nuclear buildup reflects broader trends. China’s renewable surge—led by solar and wind—has propelled non-fossil sources to over 60% of new capacity additions. Yet coal remains a backbone, providing dispatchable power to support intermittent renewables and surging demand from industry and data centers alike.

In the U.S., the picture is more challenging. The Magnificent Seven technology giants—Apple, Microsoft, Nvidia, Google parent Alphabet, Amazon, Meta, and Tesla—are driving much of the AI boom, with their data centers projected to consume power equivalent to tens of millions of homes by 2030. Forecasts suggest U.S. data center demand could more than double by 2035, accounting for up to 9% of national electricity use. To meet this, some executives and analysts speculate that these firms may increasingly resemble power companies themselves, investing directly in generation assets or co-locating data centers with new plants.

Deals with utilities for dedicated power supplies are proliferating, and tech firms are exploring onsite generation, including natural gas, renewables, and even restarts of retired nuclear units. Microsoft, for instance, has pursued nuclear revival options, while Amazon has invested in data centers adjacent to existing plants. Such moves highlight the private sector’s agility, but they also underscore grid constraints: transmission bottlenecks, permitting delays, and rising costs risk slowing U.S. AI progress.

The implications are profound. Without swift action to expand generation and modernize infrastructure, the U.S. could cede ground in AI leadership. Policymakers must streamline approvals for new nuclear, renewables, and gas projects while incentivizing private investment. China’s state-directed energy buildup offers a cautionary model of speed and scale; America can counter with innovation and market forces.

The AI race is not just about algorithms or chips—it’s about who can power them reliably and affordably. On current trajectories, China holds a commanding lead. The U.S. must act decisively to close the gap.