Nuclear power ‘essential to winning the AI race,’ industry says while arguing for regulatory reform

Industry representatives told lawmakers Thursday that revamping the regulatory environment for nuclear energy is essential to fueling power-hungry data centers in the AI era, while an expert testified that advanced nuclear reactors may carry the “potential for additional risks” and additional clean-energy sources should be part of a comprehensive strategy to meet escalating demand.

The 2024 Report on U.S. Data Center Energy Use from the Lawrence Berkeley National Laboratory found that data centers consumed about 4.4% of total U.S. electricity in 2023 and are projected to account for about 6.7% to 12% of use by 2028. A January U.S. Energy Information Administration forecast predicted electricity demand in the commercial sector to increase by 2% in both 2025 and 2026 as data-center power consumption increases.

A report last fall from consulting firm McKinsey & Company concluded that “the power needs of data centers are expected to grow to about three times higher than current capacity by the end of the decade, going from between 3 and 4 percent of total U.S. power demand today to between 11 and 12 percent in 2030.”

House Science, Space and Technology Energy Subcommittee Chairman Randy Weber (R-Texas) argued that nuclear power has “emerged as the ideal energy source” to meet this mounting demand “given its clean baseload power and unmatched reliability.”

“In addition to investments and partnerships, tech firms are also redesigning interconnection agreements,” Weber said. “Traditionally, data centers relied on front-of-the-meter interconnection agreements, in which a utility generates, transmits and distributes energy to the end user. Now, tech companies are pursuing behind-the-meter agreements, where energy assets directly power the data center.”

Subcommittee Ranking Member Deborah Ross (D-N.C.) noted that her home state has seen dramatic growth in data center construction. Amazon announced this month that it’s planning to spend $10 billion on new data centers in North Carolina.

“Much like the promise of AI, these data center investments can accelerate our economy and provide workforce benefits for years to come – but not without a cost. There are many impacts to consider in trying to meet this energy demand,” she said, citing greenhouse gas emissions and costs to ratepayers. Nuclear energy has “immense promise to be sure, but with a very high risk of cost and schedule overruns thus far.” Ross offered as an example the V.C. Summer project to build two nuclear reactors in South Carolina; abandoned in 2017 without any energy generated, the project resulted in a $9 billion bill to ratepayers.

Kathleen Barrón, executive vice president and chief strategy and growth officer at Constellation Energy, the largest owner and operator of commercial nuclear plants in the country with 21 reactors, argued that “both existing reactors and advanced plants are essential to winning the AI race.”

“America’s 94 nuclear reactors had a capacity factor of over 93 percent in 2024. These plants run 24/7, rain or shine, providing the power quality that data centers require. Nuclear plants run continuously between refueling outages, which occur every 18 to 24 months,” Barrón told lawmakers in prepared remarks. “Because these plants run around-the-clock, a 1,000 megawatt nuclear reactor provides as much energy as 2,800 megawatts of wind or 7,300 megawatts of solar, and it delivers that power every minute of the day.”

Constellation announced in September that the company is restarting Unit 1 of the Three Mile Island plant, which was shut after a 1979 accident, as the Crane Clean Energy Center in a 20-year power purchase agreement with Microsoft. The plant is expected to go back into service in 2027 or 2028. 

Earlier this month, Constellation also announced a 20-year power purchase agreement with Meta in central Illinois. “Supportive tax policy from the federal government completely changed” the outlook for nuclear power within the past few years, Barrón said.

While opining that the “most logical place for new nuclear plants is sites with existing reactors,” Barrón noted that the DOE “is also working to solve technical, operational, and regulatory challenges for a handful of advanced reactor technologies” as the Nuclear Regulatory Commission has approved some designs of small modular reactors.

“Some have suggested that federal and state policymakers require data centers to adopt a ‘bring your own generation’ policy under which technology companies would be required to build new generation before developing their data centers,” she said. “Every jurisdiction that has considered this policy thus far has rejected it, for good reason. Such a policy would effectively delay AI efforts in the United States by three to five years or more and essentially hand the AI race to China.”

Barrón advocated lowering regulatory barriers, stating that “data center projects are being stalled because of a lack of clear RTO rules for co-location, which is solely within FERC’s jurisdiction.”

“Environmental Protection Agency regulations also limit the ability of data centers to use their on-site backup generation in ways that would enable the flexibility that I mentioned earlier. Those rules currently allow backup generation to only be used in grid emergencies or a small number of hours for testing,” she told lawmakers. “Allowing those generators to be used to avoid grid emergencies in the first place would unlock the ability of data centers and other large uses of power to help the grid instead of burden it.”

Jeremy Renshaw, executive director of AI and quantum at the Electric Power Research Institute (EPRI), which recently launched the Open Power AI Consortium, said that while forecasts of power needs are “reliant on a large number of assumptions about future needs and capabilities” AI will need “support from various energy sources and continued growth in data center utilization.”

“While data centers represent a significant new load, especially in certain regions of the world, they also offer unique opportunities,” he said in prepared remarks, including workload flexibility, backup generation as a grid resource, and flexible generation and storage.

“Many perceive data centers to be power hungry, straining the grid and on an unsustainable growth pattern. Through initiatives like the Open Power AI Consortium, DCFlex and other global initiatives, we can change perspectives to utilize data centers as a knowledge factory that can accelerate knowledge generation, productivity and innovation across all sectors while growing in a sustainable fashion using advanced forms of energy generation and utilization,” Renshaw told lawmakers. “This is where advanced nuclear and other energy generation technologies may be able to support data center energy consumption needs.”

Advanced nuclear reactors may offer enhanced flexibility, improved safety, fuel diversity and higher efficiency, he said, but may also pose challenges.

“Many advanced nuclear designs have yet to be built, posing the potential for additional risks related to first-of-a-kind construction delays and cost overruns,” Renshaw said. “In addition, while used fuel inventories are generally reduced using advanced designs, used nuclear fuel and irradiated materials will still need to be managed. As with any technology, first-of-a-kind implementations may also introduce additional risks that are difficult to predict and quantify.”

“In addition to traditional and advanced nuclear, other clean, reliable generation sources and energy storage sources systems can be considered to further augment overall energy system flexibility, reliability and affordability,” he added.

Pat Schweiger, chief technology officer at small modular reactor and nuclear recycling firm Oklo, called the power demand sparked by AI “a Sputnik moment” that has fueled interest in nuclear energy. Fast reactors such as those built by his company are “walk-away safe and can be sited in closer proximity to populated areas — crucial locations for data centers and other AI infrastructure,” he said in prepared remarks.

Schweiger recommended that Congress push the Department of Energy to “accelerate its support for the domestic fuel supply chain” and high-assay low-enriched uranium; investment in research programs in the fuel cycle, commercial fuel recycling and next-generation core technologies; and reforming regulations “so that American nuclear plants can serve energy needs.”