Small modular reactor

Based on Wikipedia: Small modular reactor

In the race to power artificial intelligence, tech companies are making a surprising bet: nuclear reactors small enough to ship on a truck.

These aren't the hulking concrete domes you picture when someone says "nuclear power plant." Small modular reactors, or SMRs, are compact power stations designed to be built in factories, loaded onto flatbed trucks or barges, and assembled on-site like sophisticated LEGO sets. The largest ones produce about 300 megawatts of electricity—enough to power a small city, but a fraction of what conventional nuclear plants generate.

The smallest designs? Just 5 megawatts. That's roughly enough for a remote mining operation, a military base, or yes, a data center hungry for clean, constant power.

A Technology Born Underwater

The idea of shrinking nuclear reactors isn't new. The United States Navy has been running compact reactors since the 1950s, powering submarines and aircraft carriers with atomic energy. The USS Nautilus, launched in 1954, proved that a nuclear reactor could be made small enough to fit inside a submarine's hull while still generating enough power to propel it around the world without refueling.

But naval reactors and commercial SMRs are fundamentally different beasts. The Navy historically used highly enriched uranium—the kind that's much closer to weapons-grade material—because space on a submarine is precious, and highly enriched fuel packs more energy into a smaller volume. Commercial SMRs, by contrast, use low-enriched uranium, the same type found in conventional nuclear plants, which is far less problematic from a weapons proliferation standpoint.

Naval reactors also need to perform a peculiar trick: delivering sudden bursts of power. When a submarine captain orders "all ahead flank," the reactor must instantly ramp up to provide maximum propulsion. Commercial power plants do the opposite. They hum along at steady output for years at a time, providing the consistent "baseload" power that keeps lights on and factories running.

The Navy's six-decade safety record is remarkable. According to Admiral Frank Bowman's 2003 congressional testimony, the United States has never experienced a reactor meltdown or radioactive release from its naval nuclear program. But achieving that record required sacrificing some safety systems—there simply wasn't room for them inside a submarine hull. SMRs, with the luxury of being built on land with acres of space available, can incorporate safety features that naval engineers could only dream about.

The Oregon State University Breakthrough

The modern SMR story begins in the late 1990s, in an unlikely place: a research university in Corvallis, Oregon.

José N. Reyes Jr. and his colleagues at Oregon State University were grappling with a fundamental problem. Nuclear power was safe, clean, and efficient, but it had become impossibly expensive. After the 1986 Chernobyl disaster in the Soviet Union—which spread radioactive contamination across much of Europe and forced the permanent evacuation of an entire city—the nuclear industry had essentially frozen in place. Funding dried up. Construction projects were abandoned. Public fear made new plants politically impossible in many countries.

Reyes and his team asked a contrarian question: What if the problem wasn't nuclear power itself, but the way we were building nuclear plants?

Conventional reactors had grown enormous because of something economists call "economies of scale"—the principle that bigger factories tend to produce goods more cheaply per unit. A reactor that generates 1,000 megawatts doesn't cost twice as much to build as one generating 500 megawatts. So utilities kept building bigger and bigger plants.

But this logic had run into a wall. Nuclear plants had become so large and complex that each one was essentially a custom construction project. The sheer scale meant that components had to be built on-site rather than in factories. Construction took a decade or more. Costs ballooned unpredictably. Financing became nearly impossible.

The Oregon State researchers proposed flipping the economics. Instead of economies of scale, they would pursue economies of mass production. Make the reactors small enough to build in factories, standardize the designs, and manufacture them the way Boeing builds aircraft. The cost savings from factory production might offset the higher per-megawatt cost of smaller reactors.

By 2007, the team had invented what would become the first commercial SMR design. In 2006, they partnered with a company called NuScale Power to commercialize the technology. In 2013, they built a full-scale prototype. And in 2022, NuScale received something that had never existed before: Nuclear Regulatory Commission design certification approval for a commercial small modular reactor in the United States.

How SMRs Work

At their heart, SMRs work the same way as any nuclear fission reactor. Atoms of uranium split apart, releasing enormous amounts of energy as heat. That heat boils water into steam. The steam spins a turbine. The turbine generates electricity. It's the same basic process that coal and natural gas plants use—just with atomic energy as the heat source instead of burning fuel.

The differences lie in the engineering details, and those details matter enormously for safety.

Most SMR designs currently under development use pressurized water reactors, or PWRs—the same technology that powers most conventional nuclear plants. Water serves double duty in these designs: it carries heat away from the reactor core, and it acts as a "moderator" that slows down neutrons to the speed where they're most likely to split more uranium atoms and sustain the chain reaction.

But the SMR landscape is far more diverse than that. Some designs use molten salt as a coolant instead of water. Others use helium gas, which can directly drive a turbine without the intermediate step of boiling water. Still others use liquid metals like sodium or lead-bismuth alloys, which remain liquid at high temperatures and can transfer heat more efficiently.

Some SMRs are "fast reactors" that don't slow down their neutrons at all. These designs can be configured as "breeder reactors"—a remarkable type of nuclear plant that produces more fuel than it consumes. A breeder reactor surrounds its core with a blanket of uranium-238, the most common isotope of uranium but one that doesn't easily sustain a chain reaction. Fast neutrons from the core strike the uranium-238 atoms and transform them into plutonium-239, which is fissile and can be extracted and used as fuel. It's nuclear alchemy, turning relatively common material into reactor fuel.

The Safety Question

Every discussion of nuclear power eventually comes down to safety. The specter of Chernobyl, Three Mile Island, and Fukushima haunts the industry.

SMR proponents argue that smaller reactors are inherently safer, and the physics supports this claim to a point. The fundamental danger in a nuclear accident is "decay heat"—the heat that continues to build up inside a reactor even after the chain reaction has stopped. Radioactive isotopes created during normal operation continue to decay, releasing energy. If this heat isn't removed, the reactor core can melt.

At Three Mile Island in 1979 and Fukushima in 2011, the problem wasn't that the chain reaction couldn't be stopped—it was that cooling systems failed afterward, and decay heat melted the fuel. Because SMRs operate at lower power levels, they produce proportionally less decay heat. There's simply less energy to manage in an emergency.

Many SMR designs incorporate "passive safety systems" that work without electricity or human intervention. Instead of powered pumps circulating coolant, they rely on natural convection—hot water rises, cool water sinks—to keep the reactor cool. Instead of electronic control systems, they use physics: negative temperature coefficients in the fuel mean that as the reactor gets hotter, the chain reaction naturally slows down. The reactor essentially throttles itself.

Some designs go further, placing the reactor underground or underwater, surrounded by enormous quantities of material that can absorb decay heat. The reactor module itself might be immersed in a pool of water large enough to absorb heat for days or weeks without any active intervention.

Critics raise counterarguments. If the world needs thousands of SMRs to match the output of today's large reactors, doesn't that multiply the opportunities for accidents? Each reactor requires fuel deliveries, increasing the transportation of radioactive materials. Each reactor eventually produces radioactive waste that must be stored for thousands of years. And some SMR developers are lobbying for reduced regulatory requirements—smaller emergency planning zones, fewer backup safety systems—which could undermine the safety advantages of the technology.

A 2022 report from Germany's Federal Office for the Safety of Nuclear Waste Management took a measured view. SMRs "could potentially achieve safety advantages" through simpler designs and passive systems, the report concluded, but it's "not possible to state that a higher safety level is achieved by SMR concepts in principle." Safety depends on the total package: technology, regulations, oversight, and operational practices.

The Economics: Still Unproven

The economic case for SMRs rests on a transformation that hasn't happened yet.

The promise is straightforward: manufacture reactors in factories the way we manufacture aircraft or automobiles, standardize designs so that the hundredth unit costs far less than the first, and make nuclear power affordable for utilities that can't commit billions of dollars to a single enormous project.

The reality is messier. No company has yet achieved the kind of mass production that would prove the concept. The first SMRs are likely to cost more per megawatt than conventional large reactors, not less. The economic advantage only materializes if manufacturers build many units of exactly the same design—and the nuclear industry has historically struggled with standardization. Every utility wants custom modifications. Every regulator has different requirements. Every site has unique characteristics.

Several economic studies have found that SMR costs are "comparable" to large reactors—which is a diplomatic way of saying the expected savings haven't materialized in the projections. Critics argue that modular construction will only pay off if manufacturers receive orders for dozens or hundreds of identical units. That requires a market transformation that hasn't occurred.

NuScale Power's experience illustrates the challenges. Despite being the first company to receive NRC design certification, and despite receiving substantial government funding, NuScale has struggled to complete its first commercial project. The economics of competing with natural gas and increasingly cheap renewable energy have proven difficult.

Who's Actually Building Them?

As of 2024, only two countries have operational commercial SMRs: Russia and China.

Russia took the unusual approach of putting its reactors on a barge. The Akademik Lomonosov is a floating nuclear power plant that began commercial operation in 2020, moored in Pevek, a remote town in Russia's Far East. It houses two 30-megawatt reactors based on the designs used in Russian nuclear icebreakers—ships built to smash through Arctic sea ice that would stop conventional vessels cold. The floating design solves several problems at once: the plant can be built in a shipyard and towed to where power is needed, and at the end of its life, it can be towed away for decommissioning.

China connected its first land-based SMR to the electrical grid in 2021. The HTR-PM is a pebble-bed high-temperature gas-cooled reactor—a design that differs dramatically from conventional plants. Instead of fuel rods, it uses tennis ball-sized spheres containing thousands of tiny fuel particles, each coated with ceramic layers that can withstand extremely high temperatures. Helium gas flows through the bed of pebbles, carrying heat to generate steam and electricity. The design's proponents argue it's inherently safe: even if all cooling fails, the ceramic coatings prevent the fuel from melting.

China is also building the Linglong One, a more conventional pressurized water SMR expected to start operation by the end of 2026.

Meanwhile, the rest of the world is planning and designing and licensing. As of 2025, there were 127 different SMR designs in various stages of development worldwide. Seven were either operating or under construction. Fifty-one were in licensing processes. Eighty-five more were in early discussions with potential customers.

The Data Center Connection

The technology industry's sudden interest in SMRs reflects a collision of trends.

Artificial intelligence has created an insatiable demand for computing power, and computing power requires electricity—lots of it. Training a large language model can consume as much electricity as a small city uses in a year. Running the data centers that serve AI applications to millions of users requires constant, reliable power measured in hundreds of megawatts.

At the same time, technology companies have made ambitious commitments to reduce their carbon emissions. Google, Microsoft, Amazon, and others have pledged to reach net-zero carbon emissions, some within the next decade. Solar and wind power can help, but they're intermittent—the sun doesn't always shine, the wind doesn't always blow. Data centers need power around the clock.

Nuclear power offers something that renewable energy alone cannot: carbon-free electricity that flows constantly, regardless of weather. An SMR sitting adjacent to a data center could provide exactly the kind of dedicated, reliable, clean power that AI companies need.

Several major technology companies have announced partnerships with SMR developers. The deals are speculative—most of the reactors don't exist yet—but they signal genuine interest from companies with deep pockets and urgent needs.

Beyond Electricity

Not every SMR is designed to generate electricity. Some are optimized for heat.

Many industrial processes require enormous amounts of thermal energy. Desalination plants that convert seawater to fresh water can use heat directly, either to evaporate water in a process called multi-stage flash distillation or to power pumps that force water through reverse osmosis membranes. Chemical plants need heat for industrial processes. Hydrogen production—increasingly seen as a clean fuel for transportation and industry—requires substantial energy inputs that SMRs could provide.

Oil sands extraction, which involves separating petroleum from sand using heat and steam, could theoretically use nuclear heat instead of burning natural gas. The carbon footprint of oil sands production is notoriously large; nuclear heat could reduce it substantially, though this application remains controversial—some environmentalists argue it would simply enable more fossil fuel production.

Some SMR designs can operate in "cogeneration" mode, producing both electricity and useful heat. When electrical demand is low, excess heat can be diverted to district heating systems, desalination, or industrial processes. This flexibility could improve the economics by ensuring that the reactor's thermal output never goes to waste.

The Climate Equation

The International Energy Agency has calculated that reaching global net-zero carbon emissions by 2050 will require roughly doubling the world's nuclear power capacity. Renewable energy will do most of the heavy lifting, but it won't be enough on its own. The intermittency problem is real, and the scale of energy storage required to solve it entirely with batteries remains daunting.

SMRs could play a role in this expansion, particularly in places where large nuclear plants aren't practical. Remote communities. Industrial facilities far from electrical grids. Developing countries that lack the infrastructure to finance and operate gigawatt-scale power plants.

But the math is sobering. The world currently operates about 400 large nuclear reactors. To replace their output with SMRs producing 100 megawatts each—a generous assumption—would require 4,000 small reactors. To double nuclear capacity would require thousands more. Germany's nuclear safety office warns that meeting global energy needs with SMRs could require "several thousands to tens of thousands" of units.

Building that many reactors would require a manufacturing transformation unlike anything the nuclear industry has achieved. It would require training thousands of operators. It would require disposing of waste from thousands of sites. It would require transportation networks capable of delivering fuel to remote locations worldwide.

The Nuclear Energy Agency, an international organization that promotes nuclear power, launched an initiative at the 2023 United Nations Climate Change Conference called "Accelerating SMRs for Net Zero." The goal is to foster collaboration among governments, regulators, and industry to deploy SMRs rapidly enough to make a dent in carbon emissions before 2050.

Whether that's achievable remains an open question. The technology exists in principle. The economics remain unproven. The regulatory frameworks are still developing. And the public acceptance that would enable widespread deployment varies enormously from country to country.

The Promise and the Uncertainty

Small modular reactors represent a bet on a different future for nuclear power—one where standardization replaces customization, where factory production replaces on-site construction, and where smaller scale enables rather than constrains deployment.

The technology has genuine advantages. Passive safety systems that work without electricity. Lower decay heat that makes accidents more manageable. Modular designs that could reduce construction times from decades to years. Flexibility to serve applications beyond bulk electricity generation.

But the technology also carries genuine uncertainties. The economic case depends on mass production that hasn't been achieved. The safety case depends on regulatory frameworks that are still evolving. The climate case depends on deployment at scales that would transform the global nuclear industry.

In 2010, when Energy Secretary Steven Chu called small modular reactors "America's new nuclear option," he promised they would be ready to "plug and play" upon arrival at a site. Fifteen years later, that vision remains more promise than reality. NuScale has its certifications. China and Russia have their operating plants. But the SMR revolution, if it comes, is still waiting to begin.

The coming decade will likely determine whether small modular reactors become a significant part of the clean energy landscape or remain an intriguing technology that never quite achieved its potential. For tech companies racing to power artificial intelligence, for remote communities seeking reliable electricity, and for a world trying to decarbonize its energy systems, the stakes in that outcome are substantial.