How will the United States and China power the AI race?

Kyle Chan
While the United States has the upper hand in access to cutting-edge AI semiconductors, China has a significant advantage in energy. This “electron gap” could potentially reshape the balance of compute for AI between the two countries. The United States is already facing an energy bottleneck for building new data centers, some of which require over a gigawatt of electricity—or enough to power a small city. The International Energy Agency forecasts electricity demand for data centers in the United States will more than double from 2024 to 2030, reaching 426 terawatt-hours (TWh) or roughly 9% of total electricity demand.

China’s electricity demand for data centers is also expected to more than double over the next five years, reaching around 277 TWh by 2030. Yet, this growth in energy demand for AI is unlikely to be a constraint for China, given the country’s historically rapid pace of overall energy expansion. China already generates more than twice as much electricity as the United States and has increased its total power generation by nearly 6% per year over the past decade, according to Ember data. Notably, over half of China’s electricity growth during this period has come from clean energy sources such as wind, solar, and hydropower.

Samantha Gross
Data centers for AI training and applications are much larger than those that came before, requiring as much as one gigawatt of electricity. In fact, data centers are now described by the amount of electricity they use, not the amount of computing power they contain. No location has that much electricity generation capacity online just waiting for consumers, not to mention data centers’ need for water and other resources.

The key variable for the AI industry is speed. Utilities are used to long-term planning and projects that take years. But the AI industry is in a hurry. Competition for data center locations, within countries and among nations, will depend on the speed with which infrastructure, including capacity for electricity generation and transmission, can be delivered.

In this area, China’s ability to build things quickly, without the pressures of public opposition, might be an advantage. China is also “in the habit” of building electricity generation to meet growing demand, while electricity demand in the United States is just growing now after being flat for nearly 20 years.

David Victor
Let’s keep perspective: the energy use of AI won’t, by itself, reshape geopolitical competition. AI technology, applied to the economy and the military, will have those impacts. But the energy needed for data centers is a relatively small part of the story. Parts of the United States have relatively inexpensive electricity needed for data centers. Much of China has even less expensive power. Executives at OpenAI worry about the “electron gap,” the ability to power data centers in the United States, and any hyperscaler worries about the same.

But worries from people building projects aren’t the same as geopolitical shocks. One reason I am less worried about the energy-for-AI problem than others is that many of the scenarios of rapacious energy growth for data centers are quite frothy. They’re pumped up because in the middle of the AI arms race, nobody knows how many data centers will be needed or how quickly innovation will turn AI chips from power hogs to power sippers. Many of these projects will not be needed, especially if the AI bubble bursts and clears out a lot of dead wood in the industry. The really big story in energy will be energy-saving innovation for the chips.

Kyle Chan
China is investing in energy infrastructure projects around the world, from solar plants in Saudi Arabia to offshore wind farms in Laos. At the same time, Chinese cloud service providers such as Alibaba and Huawei have been rapidly building new data centers globally and competing more directly with American cloud service providers in AI infrastructure. By offering both energy and AI infrastructure solutions, China is pursuing a complementary set of strategies for shaping the global configuration of AI development, particularly in key regions such as the Middle East and Southeast Asia.

The United States has sought to play to its own strengths when it comes to global AI infrastructure development. In the Middle East, the United States has secured deals to provide countries with advanced American AI chips to help build gigawatt-scale data centers, such as the “Stargate UAE” project in Abu Dhabi. While the United States is unlikely to match China’s ability to build new energy projects globally, the United States can use global demand for American AI chips, cloud computing services, and AI models to shape the evolution of the global AI industry. In the meantime, third countries are also likely to pursue a mix-and-match approach, such as using Chinese-supported energy infrastructure to power data centers with American AI chips.

Kyle Chan
China has used industrial policy to become the global leader in a range of clean energy technologies, including solar, wind, batteries, and electric vehicles. Through decades of investment in R&D and scaling up production, China has dramatically lowered the costs of clean energy products and made them widely available around the world. As the United States faces energy constraints in its data center buildout, low-cost, reliable Chinese solar panels and batteries may provide an option for supplemental energy supply and backup power. Unlike fossil fuels, clean energy technologies require only an upfront capital investment rather than ongoing supply chains. As a result, they are unlikely to represent a vulnerability for U.S. data centers that China could exploit in the future through export controls or other trade restrictions.

Samantha Gross
Meeting growing electricity demand for AI is not just a matter of quickly putting steel in the ground. Each potential source of new electricity has its challenges.

Natural gas looks like a logical choice in the United States, with abundant supply and low prices relative to many other markets, including China. For this strategy to be viable, however, one needed to think ahead. GE Vernova announced that its order book for gas turbines at the end of 2025 will be 80 gigawatts, a backlog that will last into 2029. The other two major turbine manufacturers, Siemens Energy and Mitsubishi Heavy Industries, are likely in the same situation.

Renewable energy is the faster option for those who didn’t foresee today’s growth in power demand and order turbines in advance. China has a surplus of solar and battery manufacturing capacity, unlike the shortage for gas turbines. Solar generation is also generally faster to site and build than natural gas and does not depend on the availability of gas supply. However, China dominates the manufacturing of solar panels and the batteries needed to store solar power.

Over the long term, new forms of zero-carbon energy, like advanced nuclear and geothermal, will come into play. The United States really is a land of abundance, in all forms of energy. But for now, reliance on Chinese supply of solar panels and batteries likely does form a point of leverage over U.S. hyperscalers. This leverage point only applies at the construction phase—once the generation capacity is in place, further inputs from China are no longer needed.

These facts set up an interesting dichotomy in the AI race: the United States largely controls the supply of the most important chips, but China largely controls the supply of renewable energy products, a key means of quickly and sustainably powering them.

David Victor
One of the most fascinating stories unfolding in the energy-for-AI landscape is how massive new electricity needs at data centers create opportunity. For China, there may be opportunities to use AI-related demands for electricity to make fuller use of power that is already oversupplied in some parts of the Chinese grid. For the United States, it’s the opportunity to build even more clean power. Across most of the U.S. power grid, the least expensive form of reliably scalable power comes from natural gas. But the biggest of the Big Tech firms, which are also in the lead on AI, are willing to pay extra for even cleaner power—solar and wind (with batteries), nuclear, natural gas with pollution control equipment, and others. Even as the federal government is flaking on climate policy, private investors are not.

Could China exploit that American interest in clean power to gain leverage? I doubt it. There’s no question that China has a big advantage in bulk clean energy technology and manufacturing. But even a dominant market position doesn’t create leverage if buyers have lots of options. If shocks in solar panel or lithium battery markets make solar more expensive—as has been happening, a bit—the data center hyperscalers will choose other options or just gas.

Kyle Chan
With proper safeguards and limitations, Chinese investment in American clean technology manufacturing could help alleviate energy bottlenecks in the United States’ AI data center buildout. Several potential arrangements could leverage Chinese capabilities in clean energy technology while minimizing risks to U.S. national security and simultaneously supporting American manufacturing.

One possible structure would involve joint ventures between American firms and vetted Chinese manufacturers, where American partners would retain a controlling stake. Another possible structure would center around licensing agreements for U.S. firms to manufacture clean energy products using Chinese intellectual property. Two areas where these partnerships could operate are solar cells and battery cells, while sensitive components such as inverters and battery management systems would be designed, produced, and installed by U.S. firms.

More generally, developing a robust technical and commercial framework for potential Chinese participation in U.S. clean energy production could create a secure path for using Chinese clean energy technology to support American progress on AI and clean tech manufacturing.

Liza Tobin
Chinese clean energy investment isn’t analogous to Japanese auto investment in the 1980s. Japanese firms brought competitive pressure but operated as market actors under the rule of law. Japan, a treaty ally, was aligned with U.S. security interests.

China is different. It harnesses “private” companies to advance the Chinese Communist Party’s goal of dominating every critical technology supply chain. Beijing’s weaponization of rare earths is merely a preview—its leverage will grow if it strengthens its foothold in U.S. infrastructure.

The “truce” negotiated by U.S. President Donald Trump and Chinese President Xi Jinping creates new opportunities for Beijing to extract concessions: Washington’s eagerness for the deal to hold makes it tempting to downplay national security risks. But Volt Typhoon, a Chinese state-sponsored hacking group, reveals Beijing’s intent to weaponize infrastructure for sabotage, and its latest five-year plan makes it clear that the truce hasn’t tamed its technology ambitions.

While Beijing dominates global clean energy supply chains, U.S. market access remains valuable leverage that Washington shouldn’t squander by allowing Chinese investment in grid-connected infrastructure. Even if strict conditions were required—transparent ownership and complete segmentation from critical infrastructure—one would have to be naive to believe Beijing would abide by them.

The proper analogy isn’t Toyota building factories in Kentucky, but inviting the USSR to codevelop nuclear technology.