Small modular reactors promise city power from a machine the size of a shipping crate

Nuclear power in the United States has barely grown in decades, held back by the enormous upfront cost and years-long licensing timelines required to build a conventional reactor. A new generation of designs is trying to solve that problem by shrinking the reactor itself, building it in a factory instead of on a construction site, and shipping the finished modules by truck or rail to wherever the power is needed. The result, at least on paper, is a machine compact enough to fit on a flatbed but still capable of powering a small city.

From 1,500 Megawatts Down to Under 300

The Nuclear Regulatory Commission and the Energy Information Administration define the category by size rather than by any single technology. Traditional large-scale nuclear plants generate between 550 and 1,500 megawatts per unit, according to the U.S. Energy Information Administration, while a small modular reactor tops out around 300 megawatts. A further subset, called microreactors, generally produces 20 megawatts or less and can run independently of the broader grid entirely, either as a standalone power source or as the anchor of a small local microgrid. The defining engineering trait across all of them is modularity: the main components are built as factory-assembled modules and shipped to the site for installation, a process the EIA says could meaningfully cut construction times compared with pouring concrete and assembling a reactor on location piece by piece.

The Economic Pitch: Modularity Over Mega-Projects

The Department of Energy frames the appeal in economic rather than purely technical terms. Its own summary of the benefits of small modular reactors lists lower initial capital investment, siting flexibility for locations that could never accommodate a traditional gigawatt-scale plant, and the ability to add capacity incrementally as demand grows rather than committing billions of dollars to a single large project years before it generates any electricity. Design variations extend the pitch further: some SMRs use the same light-water cooling as existing U.S. reactors, while others use helium gas, molten salt, or liquid sodium as a coolant, each aimed at reaching higher operating temperatures suited to industrial heat applications like hydrogen production rather than electricity generation alone.

Powering AI and Data Centers Off the Grid

That siting flexibility is exactly why SMRs have become part of the conversation around AI and data centers. Developers who cannot get a fast, firm grid connection for a large campus are increasingly looking at SMRs and microreactors as a way to generate power on-site, bypassing the multi-year interconnection queues that have become one of the biggest obstacles to building new computing capacity. The same logic applies to remote communities and facilities with high transmission and distribution costs, where running new power lines is far more expensive than building a small reactor close to the load itself.

Fuel, Funding, and a Pipeline of Vendors

Federal money is beginning to back that shift at scale. The Department of Energy reissued a tender in March 2025 for $900 million in funding aimed at unlocking commercial deployment of American-made SMRs, and in June 2025 it launched the Energy Reactor Pilot Program, designed to fast-track testing of advanced reactor designs at sites outside the national laboratories while applicants fund their own individual pilots. Nine companies were selected for that program, including Oklo, Terrestrial Energy, Radiant Industries, and Last Energy, spanning several of the coolant technologies the EIA catalogued, from light-water to molten-salt designs.

Fuel choice is another dividing line among the designs moving through development. Several SMR concepts rely on high-assay low-enriched uranium, known as HALEU, which is enriched between 5 percent and just under 20 percent uranium-235, compared with the below-5-percent fuel used in nearly all currently operating U.S. reactors. The higher enrichment allows a reactor core to run longer between refuelings and burn its fuel more completely, which the EIA notes can shrink a reactor’s physical footprint and reduce the volume of spent fuel it eventually produces. Some high-temperature gas designs instead use tristructural isotropic, or TRISO, particle fuel, engineered to withstand extreme heat without melting down even under accident conditions, a safety feature regulators have paid close attention to as they evaluate designs that look nothing like the pressurized water reactors most of the existing fleet is built around.

The Military Is Moving First

The U.S. military has moved even faster than the civilian commercial market. A Defense Innovation Unit program launched in 2024 by the Army and Air Force named eight eligible vendors for microreactors at domestic installations, and in October 2025 the Army announced the Janus Program, which builds on the earlier Project Pele transportable reactor effort and has already identified nine candidate bases, including Fort Bragg, Fort Campbell, and Joint Base Lewis-McChord. The Air Force’s first planned nuclear microreactor is set for Eielson Air Force Base in Alaska, a pilot project with Oklo’s sodium-cooled Aurora design intended to deliver between 1 and 5 megawatts of commercially owned and operated power by 2027. The Navy, which has run reactors on submarines and aircraft carriers since the 1950s, is separately soliciting offers for commercial on-site SMRs and microreactors to power its own installations, treating the technology less as a research curiosity and more as a near-term answer to base energy resilience.

None of this erases the industry’s history of expensive setbacks. Cost overruns and schedule delays sank or scaled back several earlier SMR ventures even after they cleared regulatory hurdles, which is part of why the Department of Energy’s own pitch for the technology leans so heavily on modular manufacturing and repeatable factory production rather than one-off construction as the fix. Whether that promise holds will depend on whether the current wave of vendors can actually build reactor number ten as cheaply and as quickly as reactor number one, something none of the designs now moving through federal pilot programs have yet had the chance to prove at scale.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.

More from Morning Overview