Small modular reactors (SMRs) could become a key source of clean electricity for AI data centers, according to a major analysis published by Nature, but the technology still faces unresolved challenges in manufacturing, fuel supply, and regulatory approval. The report, released on August 26, presents SMRs as a potential solution to the surging power demands of artificial intelligence infrastructure, while spelling out the practical and political barriers that must be overcome before they can operate at commercial scale.

AI data centers need electricity around the clock, and developers are increasingly searching for reliable, low-carbon power sources located near their facilities. Conventional nuclear reactors can deliver large amounts of steady electricity, but they take years and billions of dollars to build. SMRs promise a different model: smaller units manufactured in factories with components that can be assembled on-site, similar to connecting prefabricated building blocks.

What Makes Small Modular Reactors Different from Traditional Nuclear Plants?

The scale difference is significant. An SMR can produce up to 300 megawatts of electricity, compared to roughly 1,000 megawatts for a conventional light-water reactor. This smaller footprint could make SMRs more practical for remote industrial sites or large data centers where a full-sized nuclear plant would be overkill. However, this advantage comes with a tradeoff: developers may need to build and operate many SMR units before the economics become competitive.

The U.S. Department of Energy is backing this approach with substantial funding. In March 2025, the agency opened a $900 million solicitation to support deployment of American-made Generation III+ light-water SMRs. Additionally, the DOE awarded $2.7 billion to three companies in January for domestic high-assay low-enriched uranium (HALEU) enrichment capacity, addressing one of the most critical bottlenecks in the supply chain.

How to Understand the Key Barriers to SMR Deployment

• Fuel Supply: Some SMR designs, including Kairos Power's Hermes reactor, require TRISO fuel enriched to roughly 15 to 20 percent uranium-235, known as HALEU. The United States is still building the manufacturing infrastructure needed to produce this fuel at commercial scale, creating a significant constraint on deployment timelines.
• Regulatory Approval: While the U.S. Nuclear Regulatory Commission approved a rule in 2023 allowing emergency planning zones for SMRs to be set on a case-by-case basis, each new reactor design must still prove it can operate safely and that its fuel and waste can be managed responsibly. Speed cannot replace thorough safety review.
• Manufacturing and Economics: SMR companies need factories, fuel suppliers, and paying customers, not just working reactor designs. The economics of building multiple smaller units must compete with conventional large reactors and natural gas plants, especially when data center developers want electricity on a fixed schedule.

Kairos Power's Hermes program offers a concrete test of this model. The company is constructing Hermes 1 in Oak Ridge, Tennessee, and began building the 50-megawatt Hermes 2 demonstration plant in April. Nature reports that Kairos expects commercial operations in 2030 and has signed a power purchase agreement with Google. This represents a proposed commercial pathway, but it is not evidence that a commercial SMR fleet is already operating.

TerraPower's Natrium reactor, currently under construction in Wyoming, represents a different design approach. The 345-megawatt plant uses uranium-zirconium alloy fuel, liquid sodium coolant, and molten sodium chloride for heat storage. This design sits slightly above the typical 300-megawatt SMR threshold, illustrating that engineering and commercial plans do not always fit neatly into standard categories.

Why Does Waste Management Still Matter for SMRs?

Long-term nuclear waste storage remains a politically contentious issue, even though the engineering challenges are better understood than public debate often suggests. Nature's analysis emphasizes that waste management cannot be sidestepped, regardless of how pressing the need for AI power becomes. The regulatory process must ensure that new reactor designs are reviewed thoroughly without treating speed as a substitute for evidence.

The immediate watch points are both physical and bureaucratic. Industry observers should track whether Hermes 2 advances from construction toward its 2030 operating target, whether HALEU suppliers can deliver fuel at commercial scale, and whether regulators can review new reactor designs without compromising safety standards. AI data center demand may create customers for nuclear power, but it cannot by itself resolve the cost, licensing, fuel, and waste questions that Nature has documented.

Nuclear power currently supplies about 9 percent of the world's electricity, according to Nature's reporting. For SMRs to make a meaningful contribution to AI infrastructure, the industry must solve not only the engineering challenges but also the manufacturing, supply chain, and regulatory hurdles that have slowed nuclear deployment for decades. The convergence of AI's power hunger and SMR development creates an opportunity, but success is far from guaranteed.