Small Modular Reactors in 2026: Projects, Costs & Outlook

2026 is the year SMRs moved from slide decks to concrete. Here is which projects are licensed, under construction or running, what the first units cost, why AI data centers are signing up, and what this year’s research says.

Short answer: Only two SMRs sell commercial power today, both in Russia and China. China’s Linglong One is in final commissioning. In North America, four first-of-a-kind plants now have construction approval, and the earliest are scheduled to switch on around 2030. AI data-center demand has turned SMRs from a policy idea into a market, but costs for the first units remain high.

What is a small modular reactor?

A small modular reactor is a fission reactor rated at roughly 300 MWe or less, about a third of a conventional large unit. It is designed so that major components are built in a factory and shipped to site. The bet is that repeated factory production will cut cost and schedule with every unit built, as it did for aircraft and ships.

SMRs are not one technology:

Water-cooled SMRs. Some, like GE Vernova Hitachi’s BWRX-300 and China’s ACP100, are scaled-down versions of today’s water-cooled reactors.
Advanced SMRs. Others use sodium, molten salt or helium as coolant and run hotter. They often need a richer fuel called HALEU (high-assay low-enriched uranium, enriched to 5–20%).
Microreactors. Units under about 20 MWe form a third category, aimed at military bases, mines and remote sites.

SMRs in numbers: 2026 statistics

156: SMR and microreactor projects tracked worldwide, with 95 assessed in detail (NEA SMR Dashboard 4.0, September 2026).
2: SMRs in commercial operation: Akademik Lomonosov in Russia and HTR-PM in China.
9.8 GW: nuclear capacity committed by US hyperscalers across 13 deals (SMR Intel tracker, May 2026).
4: US microreactors that reached zero-power criticality by 4 July 2026 (DOE Reactor Pilot Program).

The NEA’s previous full report, published in 2025, found:

51 designs in pre-licensing or licensing across 15 countries.
Roughly 85 active talks between developers and site owners.
A fuel bottleneck: 47 of the designs it assessed, over 60%, depend on fuel forms not yet available at commercial scale.

Where the leading SMR projects stand in 2026

Darlington, Ontario. GE Vernova Hitachi BWRX-300, 4 × 300 MWe. Under construction: licensed April 2025, reactor basemat installed. Target: end of 2030.
Kemmerer Unit 1, Wyoming. TerraPower Natrium (sodium-cooled), 345 MWe. Under construction: NRC permit 4 March 2026, construction began 23 April. Target: 2030.
Clinch River, Tennessee. BWRX-300, 300 MWe. Permitted: NRC construction permit 29 September 2026. Target: early 2030s.
Linglong One, Hainan, China. CNNC ACP100, 125 MWe. Commissioning: the first-half 2026 target was missed and no fuel load has been reported. Target: around 2027.
Hermes, Tennessee. Kairos Power (molten salt), test reactor. Under construction: nuclear construction since May 2025. Target: 2027.
Wylfa, Wales. Rolls-Royce SMR, 3 units totaling 1.4 GWe. Site design: contract signed April 2026. Target: mid-2030s.
Jizzakh, Uzbekistan. Rosatom RITM-200N, 2 × 55 MWe. Under construction: first unit started June 2026. Target: 2029.

The two biggest regulatory wins of 2026

The US Nuclear Regulatory Commission issued two landmark permits this year:

Natrium, March. It approved TerraPower’s Natrium plant in Kemmerer, Wyoming. This was the first construction permit for a commercial non-water-cooled reactor in more than 40 years.
Clinch River, 29 September. It gave the Tennessee Valley Authority a permit for a BWRX-300 at Clinch River, the first US construction authorization for a commercial SMR.

The NRC finished the Clinch River review in 14 months, four months ahead of its own schedule. An early site permit TVA had already secured in 2019 helped.

China’s Linglong One: close, not finished

Linglong One has been billed as the world’s first land-based commercial SMR since construction began in July 2021. CNNC targeted commercial operation in the first half of 2026. As of September, World Nuclear News reported it was still in pre-commissioning tests, with no fuel loading, criticality or grid connection announced. Trade trackers now expect commercial operation in 2027.

Once running, CNNC says it will produce about 1 billion kWh a year, enough for roughly 526,000 homes on Hainan.

Microreactors hit their deadline

The Department of Energy challenged 11 companies to reach criticality by 4 July 2026. Four did:

Antares Nuclear went first, with its Mark-0.
Valar Atomics followed.
Deployable Energy followed.
Aalo Atomics’ Aalo-X crossed the line early on 4 July.

These were zero-power tests that produce no electricity, but they proved a faster federal authorization path works. Antares plans electricity in 2027 and military deployments by 2028.

How much does an SMR cost?

Darlington gives the most transparent numbers so far. Ontario budgets C20.9billion(aboutUS15.1 billion) for all four units, including interest and contingency. The first unit costs C7.7billion,ofwhichC1.6 billion is shared site infrastructure. The fourth unit is estimated at C$4.1 billion, about 33% cheaper than the first. That falling curve is the whole SMR economic case in one project.

Key cost benchmarks:

Darlington first BWRX-300, including shared infrastructure: C$7.7 billion.
Darlington fourth BWRX-300: C$4.1 billion.
Ontario system operator estimate for Darlington power: 14.9 cents/kWh, with federal tax credits.
Typical first-of-a-kind SMR power cost: $90–120/MWh.
Target once production matures: $60–80/MWh.
IEA cost-parity target for SMRs by around 2040: $4,500/kW in the US.

Critics point out that a 300 MW reactor at C$7.7 billion costs far more per kilowatt than gas or solar with batteries. Supporters answer that a first unit is always expensive. They argue the right comparison is other firm, carbon-free power that runs around the clock. Both points will be tested as units two to four are built.

Why AI data centers want small reactors

Large AI campuses need hundreds of megawatts, every hour of the year, often in places where the grid is already full. SMRs offer firm output at capacity factors above 90%, a small footprint, and the option to sit next to the load.

By May 2026, Microsoft, Google, Amazon and Meta had signed 13 nuclear deals totaling about 9.8 GW, according to the SMR Intel tracker. That total mixes existing plants, restarts and new SMRs:

Meta: up to 6.6 GW. Includes TerraPower Natrium, Oklo Aurora and utility contracts.
Amazon: about 1.9 GW. A Talen Energy contract, plus a stake in up to 12 X-energy Xe-100 units.
Google: 1.8+ GW. Next-generation SMR fleet deals.
Microsoft: 0.84 GW. The Three Mile Island Unit 1 restart.

Key SMR-specific deals:

Google: signed the first corporate SMR agreement, for 500 MW from Kairos Power. Kairos’s Hermes 2 unit will feed 50 MW into TVA’s grid for Google’s data centers.
Amazon: led a $500 million funding round for X-energy.
Meta: its portfolio includes up to eight Natrium plants and a 1.2 GW Oklo campus in Ohio.
NuScale: in April 2026, agreed with Standard Power to develop two SMR-powered sites in Ohio.

Most of these deals target the 2030s, so for now many data centers bridge the gap with on-site gas.

Quick math: how many SMRs would a data center need?

Take a 200 MW hyperscale data center with a PUE (cooling overhead) of 1.2:

Total demand is 240 MW, or about 2,102 GWh a year.
One 300 MWe SMR at a 92% capacity factor covers that load.
Operators would typically add a second unit, or grid and battery backup, to cover refuelling outages.

What 2026 research says

Academic work this year moved from asking whether SMRs can power data centers to asking under what market and policy conditions they pay off.

1. Techno-economic boundary analysis of SMR cogeneration for hyperscale data centers
arXiv 2608.10999, August 2026. Li, She and Zhang (UT Dallas).

The team modelled a 200 MWe data center in the Texas ERCOT market. They ran 109 scenarios varying capital cost, power prices, policy, financing and cooling efficiency. The SMR supplies both electricity and steam for absorption chillers. In the grid-only baseline, the facility spends about $75.9 million a year, 93% of it on electricity. The headline finding: at mid-range costs, a carbon price of about $53–64 per tonne of CO₂ is enough to make the SMR option break even.

2. The role of small modular reactors in powering data centers for a sustainable digital economy
Cell Reports Sustainability, 2026. Jacob, Li et al. DOI 10.1016/j.crsus.2026.100820.

This perspective argues that SMR-powered campuses should be designed as integrated energy systems. Reactor heat, cooling, storage and computing loads would be controlled together. It treats data centers as multi-energy cyber-physical systems rather than simple electrical loads.

3. Dynamic stability assessment of grid-connected data centers powered by SMRs
arXiv 2603.09110, March 2026, presented at IEEE PES General Meeting 2026. Badakhshan et al.

A dynamic model pairs an SMR with a battery system to serve a grid-connected data center and compares it with a plain grid connection. The work addresses a practical concern: AI training loads swing quickly, and reactors prefer steady output. Batteries act as the shock absorber between the two.

4. Challenges of SMRs: cost, waste and decommissioning
Progress in Nuclear Energy, vol. 190 (2026) 105989. Kim and Macfarlane.

This is the sceptical counterweight. Reviewing US designs including the BWRX-300, NuScale module and Xe-100, the authors find wide uncertainty in levelized cost estimates. They conclude that non-water-cooled designs carry larger uncertainties in waste management than conventional light-water reactors.

India’s SMR push

India passed the SHANTI Act in December 2025, ending a state monopoly that dated to 1962. It allows private companies into plant operation, power generation and parts of the fuel cycle, under regulatory oversight.

The country targets 100 GW of nuclear capacity by 2047 and allocated ₹20,000 crore in the FY 2025–26 budget for SMR development. The Bhabha Atomic Research Centre is developing three designs:

The 220 MWe Bharat Small Modular Reactor (BSMR-200).
A 55 MWe SMR-55.
A high-temperature gas reactor for hydrogen.

Tarapur in Maharashtra has been approved as the site for both the BSMR-200 and SMR-55. The government aims to have five indigenous SMRs operating by 2033.

Pros and cons of SMRs

Benefits:

Firm, carbon-free power above 90% capacity factor.
Small footprint; can sit beside a data center or an old coal plant.
Passive safety features that cool the core without pumps.
Costs should fall unit by unit (Darlington: 33% lower by unit 4).
Faster licensing in 2026: an NRC permit in 14 months.

Drawbacks:

First units cost far more per kW than gas or renewables.
No Western SMR has sold a kilowatt-hour yet.
Many advanced designs need HALEU fuel, which is in limited supply.
Waste and decommissioning plans are less certain for new designs.
Over 150 competing designs split the supply chain.

Sources

NEA: SMR Digital Dashboard, fourth edition (Sep 2026). oecd-nea.org/jcms/pl_122015
NEA: SMR Dashboard third edition findings. oecd-nea.org/jcms/pl_108268
NRC: Construction permit for TVA Clinch River SMR (Sep 2026). nrc.gov/about-nrc/news-releases/2026/26-079
World Nuclear News: First BWRX-300 construction permit in the US.
ANS Nuclear News: NRC approves TerraPower construction permit (Mar 2026).
ANS Nuclear News: TerraPower begins Natrium construction (Apr 2026).
World Nuclear News: How Darlington’s C$20.9 billion is calculated.
Government of Canada: Darlington New Nuclear Project.
Globe and Mail: Darlington SMR cost.
Nuclear News Network: Linglong One project tracker.
World Nuclear News: First criticality under DOE pilot program.
Neutron Bytes: Four microreactors reach criticality (Jul 2026).
ANS Nuclear News: Rolls-Royce SMR contract for Wylfa (Apr 2026).
Uzbekistan RITM-200N SMR project.
SMR Intel: Every nuclear data center deal (2026).
Cor Advisors: SMR nuclear power for data centers (Aug 2026).
IEA: The Path to a New Era for Nuclear Energy.
Energy Solutions: SMR LCOE benchmarking 2026.
Li, She and Zhang, arXiv 2608.10999 (2026).
Jacob et al., Cell Reports Sustainability (2026).
Badakhshan et al., arXiv 2603.09110 (2026).
Kim and Macfarlane, Progress in Nuclear Energy 190 (2026).
World Nuclear News: India’s SHANTI Bill.
NewKerala: India targets five small reactors by 2033.