China energy overdrive proves Musk right on the AI gap

China’s massive drive to grow its energy capacity is strengthening its position in the global AI race, reinforcing concerns voiced by tech leaders such as Elon Musk and Nvidia CEO Jensen Huang.

New energy data suggests that China’s electricity build-out is accelerating rapidly and is large enough that the United States is struggling to keep pace. This divide could determine which country takes the lead in the next phase of AI development. 

At the World Economic Forum in January, Musk said the biggest constraint facing AI growth is no longer chips but electricity. He warned that soon the world could be making more AI chips than it has the capacity to power, and China would be the exception and the model maker, as it quickly expands its electricity-generation capacity. 

There is a similar argument made by Jensen Huang, which claims that AI competition exists on five levels: energy, chips, infrastructure, models, and applications. Power is at the base of that stack. Huang noted that, as a country, China already has roughly twice the energy capacity of the United States, structurally enabling it to meet growing electricity demands for AI systems.

Grid limits and delays slow US data center growth

In the United States, this strain is already visible. According to BloombergNEF, from 2024 to 2030, data centers will account for approximately 38% of the growth in US electricity demand. Data centers will consume about 7% of the country’s electricity by the end of the decade, up from current levels. 

China’s situation appears very different. Data centers are projected to account for only about 6% of the demand growth and roughly 2% of total electricity use by 2030. That doesn’t mean China’s AI sector is small. Rather, it reflects the extent to which China’s overall electricity system is much larger and more diversified, with demand heavily driven by industry, manufacturing, and electric vehicles. 

The US power sector has been unable to keep up after nearly two decades of flat electricity demand that persisted into the early 2020s. A surge fueled by artificial intelligence has inspired plans for new gas-fired power plants, but these projects can take years to construct. Developers must navigate complex regulations, lengthy approval processes, and supply-chain bottlenecks. 

Political choices also play a significant role in shaping the energy landscape, as federal-level opposition to renewable energy has delayed or cancelled clean power projects that could have helped supply electricity to the rapidly growing number of data centers. According to Michael Davidson, an energy policy expert at the University of California, San Diego, the US is effectively holding itself back by not making it easier to scale up renewable energy quickly enough to meet rising demand.

China’s rapid build-out doesn’t guarantee AI leadership

China, by contrast, continues to maintain a rapid pace of power additions across many sources. It is amassing enormous amounts of solar and wind capacity, in addition to the coal, nuclear, and gas plants. China’s entire solar capacity will overtake coal for the first time this year, even though renewable plants tend to operate with lower reliability than fossil-fuel plants. 

The connection between new data centers and the grid is much simpler in China. For most new Chinese projects, grid access is basically a “non-issue,” according to David Fishman of consultancy The Lantau Group. Goldman Sachs researchers estimate that China might have spare power capacity exceeding three times total global data center demand by 2030. 

But energy alone will not dictate the AI race’s competition. The US has a significant lead in advanced chips and AI model work. According to Google DeepMind CEO Demis Hassabis, Chinese AI companies are about six months behind the most sophisticated Western systems at the cutting edge. 

Analysts, citing insights from Gartner’s Chirag Dekate, argue that despite China’s obviously large energy potential, the US remains at the forefront of chip innovation and the building blocks for AI models. The result may depend on the US’s ability to close its energy gap quickly and whether China leverages its vast energy resources to make breakthroughs across the rest of the AI stack.

If you're reading this, you’re already ahead. Stay there with our newsletter.