Small modular reactors, explained: who is building them worldwide

Small modular reactors (SMRs) are nuclear power plants that produce up to roughly 300 MWe per module — about a third of a conventional reactor — designed so that major components can be built in factories and assembled on site rather than constructed as bespoke megaprojects. The promise is serial production: build the same unit many times, get faster and cheaper each time. Two designs already generate commercial power in China and Russia. In the West, the first construction sites are in Ontario and Wyoming.

Key facts

• The IAEA defines SMRs as reactors of up to 300 MWe per module, factory-fabricated and transportable to site
• The OECD Nuclear Energy Agency counts 127 SMR designs worldwide, 74 in active development
• Two commercial SMRs already generate power: China's HTR-PM (December 2023) and Russia's Akademik Lomonosov (May 2020)
• Only one design holds full US NRC design certification: NuScale's module, certified January 2023, with its uprated 77 MWe US460 receiving Standard Design Approval in May 2025
• The first Western SMR is under construction: GE Vernova Hitachi's BWRX-300 at OPG's Darlington site, targeting first power by the end of 2030
• TerraPower's Natrium plant broke ground in Kemmerer, Wyoming in April 2026 — the first US utility-scale advanced reactor to enter construction

How it works

A conventional reactor is a one-off civil-engineering project: a gigawatt-scale nuclear island poured and welded on site, with a safety case written for that site. SMRs invert the model. The nuclear steam-supply system is designed as a repeatable product. Heavy components — reactor vessels, steam generators, containment modules — are fabricated in a factory, shipped, and assembled on a smaller footprint. Many designs rely on passive cooling that needs no operator action and no off-site power: natural circulation, gravity-fed water, or a metal coolant that keeps transferring heat after the pumps stop.

The physics is still fission. Uranium-235 (or, in a few fast-spectrum designs, a uranium-plutonium mix) splits, heats a coolant, and that heat makes steam or drives a power-conversion loop. What changes is the coolant, the fuel form, and the size of the economic bet.

Light-water SMRs shrink proven pressurized- or boiling-water technology and keep today's low-enriched uranium fuel. That is why they lead on licensing familiarity. The BWRX-300 (GE Vernova Hitachi, 300 MWe) is a natural-circulation boiling-water reactor; Westinghouse's AP300 packages AP1000 passive safety at 300 MWe; NuScale's 77 MWe module holds the only full US design certification; Holtec's SMR-300 is in NRC environmental review for Palisades.

Advanced (non-light-water) designs change the coolant to unlock higher temperatures, storage, or fuel recycling. TerraPower's Natrium pairs a sodium-cooled fast reactor with molten-salt thermal storage. X-energy's Xe-100 is a high-temperature gas-cooled reactor firing TRISO pebbles for industrial heat as well as electricity. Kairos Power's Hermes plants use molten fluoride salt. Lead-cooled designs from Russia's BREST programme and Europe's Newcleo and Blykalla sit in the same family. These trade a well-known licence path for new capabilities — and, usually, a need for HALEU.

Microreactors are the small end of the same idea, typically up to about 10 MWe, built to move. They are aimed at a military base, a mine, a remote grid or a single data hall rather than a utility fleet. Several US test units reached first criticality in 2026 under DOE authorization; that is not the same as an NRC commercial operating licence.

The trade-off is baked in. A small reactor produces less revenue per licence, per operator, and per acre. The entire SMR bet is that repetition — the learning curve of building the same unit again and again — outruns the lost economies of scale. That bet is unproven, which is why serial programmes matter more than any single first-of-a-kind.

For a decision-oriented look at the three highest-profile Western contenders, see BWRX-300 vs AP300 vs Natrium.

The numbers

Common misconceptions

"SMRs will be cheaper than large reactors." First units are more expensive per megawatt, not less. A 300 MWe plant still needs a licence, a security force, a fuel contract and a switchyard, and it spreads those fixed costs over fewer megawatts. The cost thesis is a learning curve: factory jigs, repeatable civil works, and a supply chain that has already made the vessel once. Until Darlington's later units, Wylfa's follow-ons, or a fourteen-unit Polish order actually land, the cheaper-than-large claim is a forecast.

"No SMRs exist yet." Two commercial plants are already on the grid. China's HTR-PM has been in commercial operation since December 2023. Russia's Akademik Lomonosov has supplied heat and power to Pevek since 2020. What does not exist yet is a Western commercial SMR in power operation — and that is a narrower, more useful sentence.

"Every SMR needs HALEU." Light-water designs use the same low-enriched uranium as today's fleet. HALEU is a constraint for sodium fast reactors, high-temperature gas reactors and most microreactors, not for the BWRX-300, AP300, SMR-300 or NuScale US460. Mixing the two is how a fuel bottleneck for advanced reactors gets misread as a bottleneck for "SMRs."

"A data-center deal is a plant." The IEA's 45 GW of conditional offtake is a stack of term sheets, not steel. It shows that AI loads have pulled nuclear into procurement conversations. It does not mean 45 GW of small reactors are financed, licensed, or pouring concrete. Nature's August 2026 Outlook made the same distinction: demand is real; cost, fuel, waste and regulation are still open.

Who is building SMRs

The field is global, and the operating plants are not in North America. The OECD NEA's latest snapshot counted 127 designs, with 74 under active development. The World Nuclear Association keeps a design database and a project tracker; both show the same gap the ranking on NNN's top SMR developers is built to measure — many renderings, few construction sites. Developers with steel in the ground, a licence in hand, or a funded first plant, by headquarters:

The table is not a catalogue of every memorandum of understanding. CNEA's CAREM prototype in Argentina reached advanced civil works then stalled on funding. Last Energy, Aalo, Radiant, Antares, Valar and BWXT are moving hardware on the microreactor track — including DOE-authorized criticalities and the Army's Janus programme — but they are not yet utility plants. Buyers in Africa are sizing a $105 billion, 15 GW market by 2035; that is demand, not a named FOAK. NNN ranks the Western field separately in top SMR developers, 2026.

Current state (September 2026)

The operating scoreboard is still two plants, both outside the G7: HTR-PM and Akademik Lomonosov. The construction scoreboard is no longer a rendering. China's Linglong One is in pre-commissioning at Changjiang, with commercial operation still targeted for 2026 and not yet recorded as grid-connected in IAEA PRIS as of mid-year. Russia's BREST-OD-300 and the first export RITM-200N at Jizzakh are in civil works. In the West, Darlington unit 1 is the BWRX-300's first-of-a-kind, OPG's operating-licence application is in front of the CNSC, and TVA's Clinch River permit is in the last stretch of NRC review. Natrium is in nuclear construction at Kemmerer, with a Hyundai E&C fleet framework sitting behind the FOAK. Rolls-Royce SMR has a funded three-unit job at Wylfa and no nuclear concrete yet. NuScale finally has a project that took FID, in Romania. X-energy is waiting on a construction permit, not a design idea.

The demand signal is louder than the build signal. The IEA's 45 GW offtake figure and the Africa $105 billion market map describe buyers and governments positioning for modules that mostly do not exist yet. The next eighteen months of milestones — Linglong One grid connection, Darlington's civil progress, Natrium nuclear-island work, Clinch River's commission vote, Seadrift's permit, Wylfa site works — will separate the handful of deployable designs from the other hundred and twenty.

Latest NNN coverage

• Analysts size Africa's SMR market at $105B, 15 GW by 2035 — September 2026
• Nature Outlook puts SMRs at center of AI power debate — August 2026
• US Army picks five vendors for Janus microreactors — August 2026
• Westinghouse eVinci microreactor reaches zero-power criticality — August 2026
• TerraPower names Hyundai E&C as EPC for up to eight Natrium reactors — August 2026
• Oklo's Groves reactor reaches first criticality in Texas — August 2026
• IEA says data-center SMR deals hit 45 GW — July 2026
• Poland's Orlen Synthos seeks CfD backing for 14 BWRX-300 SMRs — July 2026
• OPG seeks licence to operate first G7 small modular reactor — June 2026

Related reading

Start with the design pages if you want hardware rather than the category: the BWRX-300, Natrium, and the side-by-side BWRX-300 vs AP300 vs Natrium. The Western ranking by construction and licences is top SMR developers, 2026. Fuel constraints for the advanced half of the field are in HALEU, explained and TRISO fuel, explained. The US construction list is every nuclear plant under construction in the US; the licence path those plants still have to walk is NRC reactor licensing, explained. Microreactor testing has its own venue in the DOME test bed.