Energy Subcommittee Hearing - Powering Demand: Nuclear Solutions for AI Infrastructure

2025-06-12

Loading video...

Summary

The subcommittee convened to discuss nuclear energy solutions for the rapidly growing demand of artificial intelligence (AI) infrastructure in the United States, examining the energy landscape and capacity to meet this need, and reviewing the Department of Energy's (DOE) programs supporting next-generation nuclear reactors [ 00:20:02-00:20:27 ]

[ 00:21:24-00:22:01 ] [ 00:27:12 ] . The discussion focused on how nuclear energy can serve as a reliable, clean baseload power source for energy-intensive AI data centers [ 00:21:49 ] [ 00:24:59 ] .

Themes

Growing Energy Demand from AI and Data Centers

The rapid growth of artificial intelligence is creating a significant demand for power, with data centers projected to consume 5.2% of U.S. electricity this year and potentially 11.7% by 2030 [ 00:23:40-00:23:54 ]

. This increase is estimated to rise from 25 gigawatts to 80 gigawatts, requiring a substantial boost in energy supply [ 00:24:04 ] . AI-specific data centers can consume up to five times more energy than traditional ones and require 24/7 supply at a massive scale due to their critical operations .

Nuclear Energy as a Solution for AI Infrastructure

Nuclear power is presented as an ideal solution due to its clean, baseload power, unmatched reliability, and high capacity factor of 92.5%, which is crucial for data centers that can tolerate minimal downtime [ 00:24:59 ]

[ 00:25:01-00:25:22 ] . Tech companies are actively investing in nuclear energy through strategic power purchase agreements and investments in nuclear developers, recognizing its importance for long-term power supply [ 00:25:34-00:25:50 ] . Additionally, the existing nuclear fleet can help meet near-term demand through life extensions, increased output, and restarting previously closed reactors . The co-location of data centers with nuclear plants is also emerging as a viable strategy to reduce transmission costs and streamline approval processes [ 00:26:31-00:26:43 ] .

Policy and Regulatory Recommendations

Recommendations for accelerating advanced nuclear deployment include continued congressional support for DOE's domestic fuel supply chain and HALU production, ongoing investment in next-generation research programs, and modernization of regulations for inherently safe nuclear technologies . Specific policy proposals include retaining Section 45U production tax credits, supporting the DOE's Loan Programs Office, and continued funding for the Advanced Reactor Demonstration Program (ARDP) . Concerns were raised about proposed budget cuts to DOE's Office of Nuclear Energy and the ARDP . State-level actions, such as Texas's Advanced Nuclear Deployment Act, are seen as vital for supporting the industry and complementing federal efforts . Streamlining licensing processes and reducing unnecessary regulatory hurdles, especially for existing nuclear sites, are also highlighted as critical steps .

Economic and Ratepayer Considerations

There is a strong emphasis on protecting ratepayers from high costs associated with new energy sources, especially given past experiences with cost overruns and delays in nuclear projects . The distinction between competitive markets, where costs are driven by competition, and monopoly markets, where regulators guarantee rate recovery, is important for understanding investment dynamics . The involvement of corporate partners in funding and guaranteeing power purchase agreements (PPAs) is seen as a way to share risk and avoid burdening ratepayers .

Innovation and Efficiency in AI and Energy Systems

Advances in chip designs and AI model architectures are improving computational efficiency, potentially blunting the growth of AI's energy needs [ 01:18:44 ]

. The concept of "Jevons Paradox" suggests that increased efficiency might also lead to expanded usage, further increasing overall energy demand . Data centers also offer opportunities for grid flexibility, such as workload shifting and utilizing backup generators as dispatchable resources, to support grid stability . Furthermore, AI itself can be used to optimize energy systems, such as hydroelectric power, leading to increased output and efficiency .

Fuel Supply and Recycling

Securing a robust fuel supply, particularly high-assay low-enriched uranium (HALU), is a critical challenge for advanced reactors . The recycling of spent nuclear fuel, leveraging past U.S. innovation, is highlighted as a promising solution to both manage waste and create a new fuel source for fast reactors .

Tone of the Meeting

The tone of the meeting was largely optimistic and urgent, underscoring the critical need for nuclear energy to meet the surging demands of AI infrastructure [ 00:21:24-00:21:31 ]

[ 00:27:12 ] . There was a strong bipartisan consensus on the importance of advancing nuclear technologies [ 01:36:25 ] . However, the discussion also carried a cautious undertone regarding the high costs and potential for schedule overruns in nuclear projects, as well as concerns about regulatory hurdles and proposed budget cuts [ 01:15:54 ] . Overall, witnesses conveyed a sense of readiness and commitment from the nuclear industry to address these challenges with strategic policy support and innovative approaches .

Participants

Transcript

Randy K. Weber Sr.

The subcommittee will come to order. The subcommittee on energy is convening. Without objection, the chair is authorized to declare recesses of the subcommittee at any time. Welcome to today's hearing entitled Powering Demand, Nuclear Solutions for AI Infrastructure. I recognize myself for five minutes for an opening statement. So listen, good morning, everybody. We're glad y'all are here. That includes the audience, by the way. Welcome to the hearing titled Powering Demand, Nuclear Solutions for AI Infrastructure. Folks, with artificial intelligence's rapidly growing demand on our power grid, this hearing is going to examine the U.S. energy landscape and our capacity to meet that need. In light of recent announcements around the country, we are going to focus on nuclear energy's role as a base load power source. We will also review the Department of Energy's research, development, and demonstration programs supporting the next generation of nuclear reactors. Much like the 1940s, most of y'all weren't here then, and just for the record, neither was I. Much like the 1940s when Americans stood at the dawn of the atomic era, we now stand on the threshold of a new age driven by artificial intelligence. With promises of increased efficiency and productivity, AI has the potential to revolutionize every single aspect of our economy and our way of life. Unsurprisingly, this potential has spurred a major influx of private capital aiming to turn these very promises into our reality. Additionally, this technology is poised to dramatically transform our electric grid and our energy sector due to the construction of those very same AI data centers. According to a recent report from a leading consulting company, data centers... Do you all say data or data?

Randy K. Weber Sr.

Data. There's one vote for data right there in the back. All right. According to one consulting company, data centers are projected to consume 5.2% of U.S. electricity this year with that share expected to increase to 11.7% by the year 2030. Their energy use would rise from 25 gigawatts to 80 gigawatts. Let's put that in perspective. One gigawatt of energy equates to roughly 294 utility scale wind turbines. 1.8 million solar panels, 103 offshore wind turbines, or one large light water nuclear reactor, which is why we're here. I see your excitement down there. Due to these immersed, immense energy demands, major technology companies and hyperscalers who traditionally sit on the energy sidelines are now climbing into the driver's seat to secure their long-term power supply. Nuclear has emerged as the ideal energy source given its clean base load power and unmatched reliability. Nuclear's capacity factor of 92.5%, let that sink in, is the highest of any energy source. This level of reliability is essential for data centers, which can afford no more than 5.25 minutes of downtime annually. No more than 5.25 minutes downtime annually. As a result, tech companies are making substantial investments in forging strategic power purchase agreements with nuclear designers. For instance, Amazon has invested over $300 million in X Energy. Google has signed a 500 megawatt power purchase agreement with Kairos Power, and Switch secured up to 12 gigawatts from Oklo Energy.

Randy K. Weber Sr.

Another witness here today, to partner with Microsoft to recommission a nuclear reactor at the Three Mile Island site. In addition to investments and partnerships, tech firms are also redesigning interconnection agreements. Traditionally, data centers relied on the front of the meter interconnection agreements in which a utility generates, transmits, and then distributes energy to the end user. Now tech companies are pursuing behind the meter agreements where energy assets directly power the data center. This approach reduces transmission costs, streamlines the approval process, and provides for greater operational control to the end user. Last year, Talon Energy sold Amazon a data center and agreed to a behind the meter power agreement supplying up to 480 megawatts from a Susquehanna nuclear power plant. Although FERC blocked this specific agreement, I believe back-of-the-meter interconnection agreements will be commonplace when we're constructing new nuclear power plants.

Sign up for free to see the full transcript

Accounts help us prevent bots from abusing our site. Accounts are free and will allow you to access the full transcript.