CLINTON, ILLINOIS – JULY 25: An aerial view shows storm clouds moving over Constellation’s Clinton Clean Energy Center’s single nuclear reactor power plant on July 25, 2025 in Clinton, Illinois. Meta recently signed a 20-year power purchase agreement with Constellation for the output from the plant. (Photo by Scott Olson/Getty Images)
Getty Images
The artificial intelligence revolution isn’t limited by chips, algorithms, or talent. It’s constrained by electricity. Hyperscale data centers already draw more power than entire cities, and with tech giants pouring billions into artificial intelligence, the question is no longer whether demand will skyrocket—it’s where the energy will come from.
Nuclear power, long treated as a “quiet contender” in the clean-energy mix, is suddenly viewed as indispensable. From small modular reactors that can sit alongside data centers to novel fuel-enrichment technologies that promise to secure America’s uranium supply, the industry sees a historic opening: becoming the backbone of the AI era.
“For an industry that requires downtime of maybe several minutes a year, nuclear’s unmatched consistency is an incredibly useful asset,” James Walker, CEO of NANO Nuclear Energy Inc., told me.
AI’s energy appetite is voluminous. The International Energy Agency projects that data centers could consume 8% of total U.S. electricity by 2030—and globally, demand may rival Japan’s entire consumption. Individual hyperscale facilities can draw as much power as a mid-size city, and the race to build hundreds more is well underway.
Meeting that demand isn’t as simple as adding wind or solar capacity. Renewables face land-use limitations and storage shortages. Natural gas, which has been the go-to source, is in a bind: long-term contracts have tied up turbine supply into the 2030s. Meanwhile, America’s aging grid is already at its maximum capacity; upgrading transmission lines and pipelines would cost billions and take more than a decade.
That leaves nuclear, which, despite its maturity, has yet to come into its own. SMRs have enormous potential, but given the regulatory difficulties and historical concerns, they must still overcome those hurdles. While some heavy-hitters are investing in the technology—think Bill Gates and the Natrium project and TerraPower—the soonest these reactors will hit is the early 2030s.
Right-sized reactors range from 50 to 300 megawatts, but modules can be combined into a 1,000-megawatt plant. If one module fails, it can be repaired while the others continue to operate.
Nuclear Is Consistently A Top Producer
VOLGODONSK, RUSSIA – MAY 16; The over 300 ton reactor pressure vessel. An inside look at the mega manufacturing facility of Atomenergomash a subsidiary of Rosatom the Russian atomic mega giant. Inside the facility that is the size of several football fields, a road runs through it and workers commute from one work station to the next using bicycle. This is the facility where Russia makes the heavy machinery needed for the 1200 MW atomic reactors it makes across the world. For each reactor some 6000 tons of equipment is shipped from this facility. The reactor vessel of the pressure vessel is over 300 tons in weight and it takes over two years to manufacture. (Photo by Pallava Bagla/Corbis via Getty Images)
Corbis via Getty Images
Unlike renewables or fossil fuels, nuclear power offers the highest capacity factor—roughly 93% uptime, compared to 60% for gas, 35% for wind, and 25% for solar. For hyperscale data centers, where even brief downtime can cause billions of dollars in lost productivity, this level of reliability is essential.
But the real breakthrough lies in where nuclear power can be delivered. Walker argues that advanced modular reactors—the SMRs— can be placed directly alongside AI data centers, forming dedicated microgrids. That bypasses the strained national grid entirely, eliminating transmission bottlenecks.
“You don’t need to send power through the grid system,” Walker explained. “You can simply co-locate in remote areas like the Arizona desert and know your power will be consistent and available wherever and whenever you need it.”
Unlike legacy plants, such as Westinghouse’s AP1000, which require large exclusion zones and gigawatt-scale builds, SMRs are designed for flexibility. They require less land, offer enhanced safety profiles, and can be scaled incrementally as data-center demand grows. It’s a model tailor-made for tech giants who want guaranteed power without waiting on a national infrastructure overhaul.
Yet even if the reactor technology works perfectly, one problem looms: fuel. Nuclear plants rely on enriched uranium, and currently, the U.S. supply chain is fragile. Roughly 40% of the world’s uranium enrichment capacity belongs to Russia, leaving American utilities exposed to rising geopolitical tension.
That’s where Jay Nu, Chairman and CEO of LIS Technologies, sees opportunity. His company is developing the only U.S.-made patented technology for laser uranium enrichment—an alternative process that aims to cut costs compared to the widespread centrifuge systems, which are currently dominated by Russia.
“Centrifuge enrichment is costly, slow, and largely controlled abroad,” Nu told me. “Laser enrichment can cut those costs by as much as 75% and allow the U.S. to rebuild its nuclear infrastructure.” It’s not just about making fuel cheaper. It’s about securing the supply chain for both today’s reactors and the advanced designs of the future.
The U.S. Department of Energy has taken notice. In December 2024, LIS Technologies was one of six firms selected for a $3.4 billion program to expand enrichment capacity over the next decade. LIS is developing a new facility at Oak Ridge National Laboratory, a site with nuclear history dating back to the Manhattan Project.
The Other Bottleneck: Fuel
A demonstration loop for both Low Enriched Uranium and High Assay Low Enriched Uranium is scheduled for the second half of next year. Some advanced reactors under development will require higher fuel enrichment levels, up to 20% U-235. This allows for more nuclear fission and generates more energy.
If successful, the technology could cut commercial enrichment costs to $25–$40 per separative work unit, compared to $100–$150 today. That would make domestic fuel much more competitive while ensuring supply for both current reactors and next-generation designs. SWU is the industry’s standard measure of the effort required to enrich uranium, essentially capturing the amount of energy and technology needed to increase the concentration of fissile uranium-235.
Two forces are converging to create this moment: geopolitics and AI. Russia’s invasion of Ukraine highlighted how vulnerable the U.S. nuclear fleet is to foreign fuel supplies. At the same time, the AI boom has underscored the criticality of electricity to the U.S. economy and especially tech giants.
Amazon has already invested $500 million to site a data campus near a nuclear facility. Microsoft and Google are openly exploring partnerships with nuclear reactors. Even Dow will develop a small nuclear unit for industrial applications. These aren’t experiments in corporate sustainability branding—they’re practical responses to the single biggest constraint on AI growth: power.
Neither Nu nor Walker suggests that nuclear alone will solve every energy challenge. But together, their perspectives highlight the two missing links in scaling nuclear for the AI age. Walker emphasizes deployment—getting reactors close to where the power is needed, without relying on a stressed grid. Nu emphasizes supply—ensuring there’s enough enriched uranium at a cost that makes reactors competitive.
One without the other falls short. Reactors can’t run without fuel; fuel is meaningless without reactors that can deliver it where it’s needed. Combined, they form a two-link chain that could shift nuclear energy from a “quiet contender” to a central pillar of U.S. energy strategy.
“Global electricity demand from data centers is set to more than double over the next five years, consuming as much electricity by 2030 as the whole of Japan does today. The effects will be particularly strong in some countries,” said the International Energy Agency’s Executive Director Fatih Birol, in a release.
For example, in the United States, data centers are expected to account for nearly half of the growth in electricity demand; in Japan, this figure is also over half.
If AI is to reach its full potential, its energy source must be equally revolutionary. That’s why America’s capacity to drive its next major technological breakthrough may depend on whether nuclear finally clears the dark clouds and steps into the light.