Will nuclear power be the answer for data centers’ voracious energy needs?
Data center expansion could increase electricity demand in the U.S. by 15-20% over the next decade, projects the Department of Energy. At their current rate of growth, data centers will be consuming 9% of the electricity generated in this country by 2030, compared to 4% today, according to estimates by the Electric Power Research Institute.
Much of that increased demand will be driven by the encroaching presence of artificial intelligence, the use of which for a simple Internet search requires 10 times more energy than a comparable Google search, said Amy Roma, a partner at the law firm Hogan Lovells, during a panel discussion hosted last October by the American Nuclear Society. At that same event, Seth Grae, CEO of the nuclear fuel technology developer Lightbridge, stated that 40% of data centers’ energy consumption went toward cooling.
These voracious energy needs, though, must be reconciled with the equally urgent need to reduce the built environment’s carbon footprint from construction and operations that accounts for 42% of CO2 emissions. That is what’s motivating Big Tech companies—which pumped more than $180 billion into data center expansions and related infrastructure in 2024, according to the research firm Dell’Oro Group—to turn to nuclear as a future alternative power source. Several companies have struck strategic investment agreements to secure their access to this cleaner power, and to support the development and potential commercialization of advanced nuclear reactors.
While Big Tech has been snatching the headlines, some leading AEC firms active in this sector are also playing roles as both service providers and partners.
• Last summer, the engineering firm Burns & McDonnell entered into an agreement with BWX Technologies to further the design and realization of microreactors (MMRs) that generate up to 30 MW of electricity. Burns & McDonnell’s project scope includes developing power cycle architecture, identifying critical components, and integrated design. The Department of Energy’s Strategic Capabilities Office selected BWX to make a full-scale mobile microreactor prototype, that was scheduled for delivery in late 2024 to be tested by the Idaho National Laboratory. Scott Strawn, Vice President and GM of Burns & McDonnell’s Power Group, said in a prepared statement his firm ‘s project with BWX would be the first domestic nuclear application that produces electricity and steam for industrial use.
• Last September, the engineering firm Jacobs Solutions completed the merger of its Critical Mission Solutions and Cyber and Intelligence businesses with Amentum, a global engineering and technology services business based in Virginia and serving the U.S. government and allied partners. Amentum has decades of nuclear construction experience; going back to the 1970s, its heritage firms helped to build every one of the United Kingdom’s nine existing nuclear power stations. More recently, Amentum is functioning as a technical and development advisor in Poland and the United Arab Emirates.
A cleaner imperative
Industry experts assert that nuclear power must be considered an option for carbon-free data center development and operation, and that smaller reactors—with their lower costs, simpler design, and quicker construction times—could work in tandem with other renewable power sources to help meet the impending demand for clean energy production.
To date, there are no microreactors or small modular reactors (SMRs), which produce up to 300 MWs of electricity, available commercially in the U.S., although more than a dozen versions are in various stages of development or review.
The think tank Third Way reported a year ago that X-energy, the nuclear reactor and fuel design engineer, has been working with the Department of Energy to construct the first commercial high-assay low enriched uranium (HALEU) fabrication facility, to ameliorate shortages in fuel procurement.
“Future data centers will require significantly different workloads and power to operate, and this level of utility is not immediately available in almost any region of the world, especially not as carbon-free power,” says Andy Solberg, Jacobs’ Global Market Solutions Director of Decarbonization. “Parallel paths are needed to generate new gigawatt-scale power.” Solberg expects MMRs to be online before 2030, and SMRs in the early to mid 2030s.
That remains to be seen, as the nuclear industry has been plagued by cost and schedule overruns. The Vogtle 3 reactor in Georgia, which came online in 2023, was seven years behind schedule and $21 billion over budget. (The fourth unit in that nuclear plant started operating a year later.) The Chicago Tribune reported that half of the 250 attempts to build nuclear reactors since 1960 were canceled before any electricity was generated.
Given that history, “the endeavor to blanket the Earth with SMRs is a Hail Mary pass that’s very unlikely to succeed,” Paul Hockenos, a writer who specializes in energy and climate reporting, posted on the website Undark.org.
Nuclear already a factor
Optimists about the future of nuclear point out that nuclear is already an undeniable power source. There are 54 nuclear power plants with 94 reactors operating in the U.S., with an aggregate capacity of 96,952 MWe. The U.S. produces 30% of the global nuclear electricity, and its reactors generated 77TWh in 2022, 18% of the total electrical output, according to statistics culled by the World Nuclear Association. Nuclear is, by far, the largest green energy source, and its Capacity Factor—which measures how often a plant is running at maximum power—is 93%, two times that of natural gas and three times that of solar and wind, according to Department of Energy estimates.
DOE projects that the U.S. will need at least 200 GW more nuclear power—three times its current capacity—to meet future electricity needs. In November, the Biden Administration unveiled its plans to triple the nation’s nuclear power supply by midcentury and offered $900 million in grants for the development of smaller reactors. Eight western states are crafting economic development plans that are focused on advanced nuclear deployment, with help from the Idaho National Laboratory’s Frontier Initiative.
Jockeying for better position
Big Tech isn’t waiting for new power sources to emerge and has been brokering deals to ensure it has access to clean power to handle its AI-infused requirements.
• Microsoft recently struck a 20-year power purchasing agreement with Constellation Energy that is predicated on the reopening of the Unit 1 nuclear power plant at Three Mile Island, which is planned for 2026. That energy will be used to power Microsoft’s data centers;
• Last October, X-energy announced a collaboration with Amazon to bring online more than 5 GW of new nuclear power plant projects—encompassing 64 of X-energy’s Xe-100 SMRs—in the U.S. by 2039. Amazon also has an agreement with Energy Northwest, a consortium of state public utilities, to enable the development of four SMRs that in their first phase would generate 320 MWs;
• That same month, Kairos Power and Google stated that they were collaborating to deploy 500 MW of advanced nuclear power by 2035, with the first deployment expected within the next five years.
“Investments from tech are already challenging traditional ideas about financing, regulating, and building nuclear energy,” stated Hogan Lovells, the law firm. For example, no nuclear plant firm that closed has ever been reopened. The law firm also noted that in December 2023, Kairos received a construction permit from the U.S. Nuclear Regulatory Commission for Kairos’ Hermes 1 prototype reactor and a second permit for its Hermes 2 reactor last November, the latter of which will be built at the Heritage Industrial Park in Oak Ridge, Tenn.
In response to written questions from BD+C, Kairos Power, which was founded in 2016, said it is focusing on the commercialization of its fluoride salt-cooled high-temperature reactor technology, known as KP-FHR. Its standard commercial offering will be a two-unit, 150 MWe plant (75 MWe/unit) with a single power conversion system. Kairos is working toward a U.S. demonstration of this technology by 2030, leading to a “rapid deployment ramp” in the 2030s.
Other SMRs and MMRs in the works (as summarized by the World Nuclear Association) include Westinghouse Electric’s modular AP-100 model, which the company claims can be built in 36 months. NuScale, based in Portland, Ore., was the first developer to have its SMR design approved by the NRC. In 2022 NuScale completed the sixth stage of design certification for its 50-MWe unit, which it was planning to test at Idaho National Laboratory. NuScale was also seeking separate approval for a 77-MWe version.
A year ago, Holtec International revealed plan to build its first two 300-MWe SMRs at its Palisades nuclear plant in Michigan, for which it intends to file a construction permit with NRC by 2026, with a target commissioning date in the mid 2030s. South Carolina Electric & Gas has been evaluating X-energy’s 50-MWe Xe-100 to replace coal-fired plants. And Last Energy continues to develop its PWR-20 microreactor.
Cost and delays still roadblocks to commercialization
Even advocates of advanced nuclear power acknowledge that the newer reactors have some roadblocks to overcome along with their advantages.
On the plus side, says Kolberg of Jacobs, are increased safety, faster construction times, higher efficiencies, less nuclear waste, and potentially lower CapEx and OpEx costs. The International Atomic Energy Agency (IAEA) adds that SMRs can be situated on locations not suitable for larger power plants, and their designs are generally simpler and don’t require human intervention to be shut down. SMRs also have reduced fuel requirements, needing to be refueled every three to seven years versus every one to two years for conventional plants. And SMRs can be paired with other renewable sources in a hybrid energy system.
The nuclear industry, though, has some perennial issues that still need to be resolved. In a white paper he posted in November 2024, Kevin Fox, an engineering manager at Burns & McDonnell in Fort Worth, cited as obstacles to commercialization fuel availability and storage, security, operations and maintenance, upfront costs, and licensing and permitting.
But most industry watchers don’t see these as insurmountable problems, and they anticipate sizable expansion in reactor demand. Reuters reported in October that 14 banks from U.S. to Asia had signed a joint declaration to increase their support for nuclear power construction. The banks' pledge comes a year after 22 countries and 120 companies signed a joint statement at the COP28 conference in Dubai to triple global nuclear capacity by 2050 as part of the drive to cut greenhouse gas emissions.
In light of these commitments, Reuters projects that the global SMR market will reach $72.4 billion by 2033, “with the recent data center energy crisis driving huge growth potential.” Reuters is conducting a conference on SMRs and advanced reactors in Nashville next May, where it expects to draw more than 600 attendees.
The news service says the “critical themes” to accelerating SMR commercialization revolve around minimizing “first-of-a-kind” risk by adopting a fleet rollout mentality; unlocking funding by demonstrating ROI, streamlining regulation and licensing, navigating technological hurdles via standardization and scaling, and expanding and strengthening supply chains and workforces.
It also bears watching whether the market shows a preference toward nuclear energy systems, which include pebble bed reactors that use tri-structural isotropic coated fuel; SMRs that use only a fraction of the uranium needed in conventional plants; or liquid fluoride thorium reactors. “We believe global demand will support a number of different vendors across many different markets,” predicts Jon Fowler, Amentum’s Vice President of Energy. Equally relevant determining factors for deployment, he says, include delivery capability, and a resilient supply chain that enables deployment across different countries.
Jacobs’ Solberg thinks that data center professionals have a “tremendous opportunity” to set a new standard for efficiency and zero-emission power. “The U.S. will meet its energy goals. The challenge is to avoid being short-sighted by hastily deploying outdated power solutions.”
