United States' Domination in Nuclear Energy and Artificial Intelligence: The Intricate Link with Data Centers
The United States is falling behind in the industrial race, surrendering grounds in batteries, EVs, solar, wind, and mineral processing to China. Yet, America maintains its top spot globally in artificial intelligence (AI), a technology considered to be among the most transformative of our times. However, holding onto this position relies heavily on a crucial element: powering the substantial data centers that drive AI advancements.
Europe appears to have conceded defeat in the innovation race. This is evident in the Draghi report and the recent industry exodus from Germany, one of the EU's powerhouses. On the other hand, the U.S. still harbors a lot of fighting spirit. An alliance between tech entrepreneurs and the energy sector could ensure both economic growth and environmental sustainability, provided it's executed wisely.
Maintaining American dominance in AI necessitates sustainable, substantial power. This is where nuclear energy proves to be a game-changer. The year 2024 heralds the beginning of a long-awaited nuclear revival, marked by three significant advancements: 14 major banks committing to finance nuclear projects, the tech sector's endeavor to reactivate retired nuclear power plants, and the signing of the ADVANCE Act, which altered the Nuclear Regulatory Commission's mission.
This revival coincides with an extraordinary time for our electric grid, which is undergoing a dramatic transformation with a steady 3% growth through 2035. Beyond load growth, we're also decarbonizing and retiring outdated generating capacity. Data centers will drive the most substantial expansion—growing from low single-digit percentages to 22% consumption—while other sectors will also witness significant increases, with transportation accounting for 46% of electricity usage, buildings 16%, and industry 15%. Meeting this escalating demand, particularly for energy-intensive data centers, necessitates power sources that can deliver both large scale and reliability.
Nuclear power plants and advanced supercomputers are a perfect match. Operating Nvidia's AI chips may require up to 1 GW of energy—precisely the amount a typical AP1000 nuclear reactor can provide. While data centers could theoretically run on solar and wind, intermittency poses a crucial challenge. Ensuring sufficient renewable electricity would necessitate a substantial overbuild of grid infrastructure, potentially up to a factor of 4 or more, alongside utility-scale battery storage. When electricity shortages occur, residential customers would be at risk of blackouts, instead of data centers. With nearly 80% of global internet traffic passing through servers in Northern Virginia, even brief disruptions could result in billions in economic damage.
Natural gas power plants offer an apparent solution, but they too face an insurmountable hurdle. Many tech companies have pledged sustainability goals, promising to use clean energy both synchronously and in the same region as their data centers. Utilizing natural gas for data centers long-term is unfeasible due to the tech sector's climate commitments, as carbon capture and storage remains excessively costly.
Nuclear power emerges as the only feasible option to meet these demands. Currently, the United States boasts the most nuclear capacity with 94 operational reactors and nearly 100 GW of capacity. Yet, it necessitates tripling this capacity by 2050 to meet growing demand, as reflected in the Department of Energy's plan for 200 GW of new nuclear capacity. However, China is poised to take the lead soon, as the U.S. has largely stopped building new plants. Since 1996, only three new reactors have been commissioned: Watts Bar 2 in TN (2015), Units 3 and 4 in Plant Vogtle in GA (2023-2024).
Question: How can the U.S. construct multiple nuclear plants annually until 2050? The nuclear industry has a saying: "France has two types of nuclear reactors and a hundred kinds of cheese, while the U.S. has two kinds of cheese but around a hundred types of reactors." Building numerous first-of-a-kind nuclear reactors has eroded industrial expertise and inflated costs. While Georgia's Vogtle plant unit 3 faced initial delays and cost overruns, the subsequent reactor proved significantly more economical—demonstrating clear learning-by-doing benefits. But focusing only on construction costs overlooks the bigger picture. Recent research suggests that current LCOE calculations might be biasing against nuclear power, as they do not account for whole-system costs, capacity factors, and reliability benefits. The true economic picture becomes clear when examining fully depreciated nuclear power plants, which can produce electricity at remarkably competitive rates around $31 per MWh.
Nuclear's high costs are not inherent to the technology. They are self-inflicted through excessive regulation. We have imposed overly stringent safety standards based on the misconception that any level of radiation is hazardous, no matter how minimal the dose. This notion defies common sense. If this were true, wouldn't we see severe health impacts in populations residing in areas with high natural background radiation? However, we do not. These misguided policies have established artificial barriers to nuclear deployment at a time when we need it most.
The strategic priority is clear: The U.S. must keep its lead in AI, establish energy self-sufficiency through nuclear power, and protect the environment while fostering economic growth. We cannot buy our way to prosperity, but we can innovate our way forward. Nuclear power ought to be perceived as a versatile base for clean, firm power generation, pink hydrogen production, and industrial heat—all while ensuring grid stability. While the Trump presidency might view natural gas as the key to energy self-sufficiency, nuclear power can fulfill the bulk of the new power demand.
American leadership in reducing data center energy demand could significantly contribute to the overall decrease in electricity use, thereby reducing the need for an increase in nuclear power. However, with the escalating demand for energy-intensive data centers, nuclear power remains a viable option due to its ability to deliver large-scale, reliable power.
The strategic alliance between the tech sector and the energy sector, in light of American leadership, could pave the way for the efficient use of artificial intelligence, nuclear power, and other clean energy sources, ultimately leading to environmental sustainability and economic growth.
