The Duck Curve

What are SMRs and microreactors?

A small modular reactor, or SMR, is a nuclear reactor built smaller than the plants in today's fleet, with as much of it as possible made in a factory. The International Atomic Energy Agency puts the ceiling at 300 MW(e) per unit, "about one-third of the generating capacity of traditional nuclear power reactors." Microreactors are the small end of the class: typically up to 10 MW(e) by the IAEA's count, and 1 to 20 megawatts of heat for most designs, according to the Department of Energy, which describes them as small enough to move by truck.

The first commercial SMR in the US now has a permit to be built. On September 28, 2026 the Nuclear Regulatory Commission gave the Tennessee Valley Authority a construction permit for a 300 MW BWRX-300 at its Clinch River site in Tennessee, a project the agency calls the nation's first commercial small modular reactor (Duck Curve coverage). Abroad, Russia's floating Akademik Lomonosov has produced power from two 35 MW(e) SMRs since it began commercial operation in May 2020, the IAEA noted in 2023. Microreactors are ahead in hardware and behind in customers. Five designs went critical in test runs in the summer of 2026 (Duck Curve coverage), yet "no nuclear microreactors are supplying power to the commercial electric grid in the United States today," the Associated Press reported in August.

Built in a factory, cooled by physics

The pitch is a trade of scale for repetition. Many SMRs "can be factory-assembled and transported to a location for installation," the IAEA says, and it sees them going where big plants do not fit: industrial sites and remote areas with limited grid capacity. DOE goes further for microreactors, saying all of their components would be assembled in a factory and shipped out, which it says avoids the difficulties of large-scale construction and reduces capital costs.

The other half of the pitch is safety that does not depend on pumps. The BWRX-300 at Clinch River removes heat from its core "by natural circulation," per the NRC's safety evaluation, and DOE says microreactor designs would use passive safety systems that prevent overheating.

Small does not always mean new. The BWRX-300 is a boiling light-water reactor: ordinary water carries the heat out of the core. Many other designs change the coolant entirely. TerraPower's 345 MW Kemmerer Unit 1 in Wyoming, just above the IAEA's SMR ceiling, is cooled by sodium, and when the NRC authorized its construction permit in March 2026, it called that the first approval for a non-light-water reactor in more than 40 years. Microreactors, DOE notes, "are not defined by their fuel form or coolant."

Many of them need a fuel that barely exists yet

Today's reactors burn uranium enriched to at most 5% uranium-235. Most US advanced reactor designs need more: high-assay low-enriched uranium, or HALEU, enriched to between 5% and just under 20%, which DOE says they require "to achieve smaller designs that get more power per unit of volume." DOE's own explainer states the problem plainly: most of these designs "will require a fuel that isn't yet available at a commercial scale."

Many microreactors also use TRISO fuel (Duck Curve coverage): particles about the size of a poppy seed, each a uranium kernel wrapped in layers of carbon- and ceramic-based material that keep fission products in. DOE says the particles cannot melt in a reactor. The catch is supply, and it shows up in the economics. A September 2026 preprint from the University of Colorado Boulder found SMRs, at their makers' advertised costs, "uneconomical primarily because investment cost reductions are offset by increased marginal costs," and says novel fuels without established supply chains "most likely ballooned" those costs (Duck Curve coverage).

Who has to say yes

Test reactor · Department of EnergyDOE authorizationAtomic Energy ActBuildat a test siteCriticality testphysics, not powerDatafor a later licenseCommercial plant · NRC, two-step routeConstruction permit18-month targetBuildat the plant siteOperating licensea separate applicationPower to the gridno US SMR there yet
Two tracks, two regulators. A commercial plant can instead take one combined NRC license, and the Army will license its Janus reactors itself.

A commercial reactor selling to the grid needs the NRC, by one of two routes. Clinch River took the two-step one: a construction permit now, and a separate operating license before it can run. Its permit says a license authorizing operation will not be issued until TVA submits its complete final safety analysis report and the permit's other requirements are met. The review took 14 months, against the 18 months the NRC now targets for construction permit decisions. The other route is a single combined license covering construction and operation; NuScale's Idaho partners were preparing one, World Nuclear News reported. The rules themselves are being rewritten: comments on the NRC's proposed reactor-licensing rule close November 9, 2026 (Duck Curve coverage).

A test reactor can skip the NRC. DOE can authorize reactors itself under the Atomic Energy Act, and in 2025 it set up a Reactor Pilot Program with the goal of getting at least three test reactors critical by July 4, 2026. It announced the third on July 1. Those were criticality tests: a fueled core brought to a self-sustaining chain reaction at negligible power, a physics checkpoint that comes long before electricity.

The military is a third track. The reactors the Army plans for its bases under the Janus program "will be licensed by the Army, rather than the U.S. Nuclear Regulatory Commission," AP reported, with the Army saying it is aligning its process so companies can later seek NRC licenses (Duck Curve coverage). As of late August 2026, the confirmed buyers of US microreactors were mostly in uniform.

The question nobody has answered yet: cost

Small reactors were supposed to dodge the cost overruns of big ones. The record so far is thin. NuScale's plan with the Utah Associated Municipal Power Systems for six modules totaling 462 MW at a site near Idaho Falls, backed by a DOE cost share of up to $1.4 billion, ended in November 2023 when the partners concluded "it appears unlikely that the project will have enough subscription to continue toward deployment," as World Nuclear News reported. The Boulder preprint's finding points the same way from the other end: at manufacturers' advertised costs, it says, prices and subsidies "beyond historic norms" would be needed to attract private investment. The projects moving now have government money behind them, from the $400 million DOE selected TVA to receive for Clinch River and later units to the Army's contracts.

What it is not

None of this is fusion, which is a different physics and a different story (#fusion). And a reactor going critical is not a reactor making power: every summer 2026 milestone was a test. Where the permits, test reactors, military programs and fuel stand today is the job of the maintained Advanced reactors in the US reference page; the news as it happens is under #smr and #microreactors.

Sources