Small modular and advanced reactors, October 2026

For about fifteen years, small modular reactors were mostly slide decks. In 2026 that changed in a measurable way: the NRC issued construction permits, concrete was poured, and several very small test reactors went critical in Idaho. None of the new commercial designs was producing power for the U.S. grid as of October 2026. This lesson gives the status of each major project with dates, so you can tell the difference between a permit and a press release.

What "small" and "modular" mean

A small modular reactor is usually defined as 300 MWe or less, against about 1,100 MWe for a large plant like Vogtle 3. "Modular" means building major pieces in a factory and shipping them, instead of assembling everything in the field. A microreactor is smaller still, from a few kilowatts up to roughly 20 MWe, often designed to fit on a truck.

The pitch: smaller units cost less each, so a utility does not bet the company on one project; factory production should cut cost with each unit built; and smaller cores can rely on passive cooling. The skeptic's reply, covered in lesson 8, is that reactors have historically gotten cheaper per kilowatt by getting bigger, and the factory savings only arrive after many orders.

The licensing steps, in order

News coverage and press releases mash these together. They are different things.

Project by project

NuScale

NuScale was first through the NRC door. Its 50 MWe design was certified in 2022, and the larger 77 MWe US460 received a standard design approval in May 2025, ahead of schedule. Its first planned plant, the Utah Associated Municipal Power Systems' Carbon Free Power Project, planned for a site at Idaho National Laboratory, was cancelled in November 2023 after the target power price rose from $58 to $89 per megawatt-hour and participating cities pulled out. In September 2025, TVA and ENTRA1 Energy (NuScale's commercial partner) announced plans for up to 6 GW of NuScale plants in TVA's service area. That was a development agreement with power purchase deals still to be signed, not a firm order.

GE Vernova Hitachi BWRX-300

This is a scaled-down boiling water reactor that leans on decades of BWR experience. Ontario Power Generation is building four at Darlington in Canada; its May 2025 investment decision was based on a forecast of C$7.7 billion (about US$5.6 billion) for the first unit, a figure worth knowing when anyone calls SMRs cheap. TVA's Clinch River unit holds the first U.S. construction permit for a BWRX-300 and for a commercial light water SMR. In December 2025, DOE selected TVA for up to $400 million in cost-shared federal funding toward it.

TerraPower Natrium

Backed by Bill Gates and sited next to a retiring coal plant in Wyoming, Natrium is the most advanced non-light-water commercial project in the country. Its HALEU fuel supply (lesson 3) delayed the original schedule. The NRC review was completed in 18 months against an initial 27-month schedule.

X-energy

Two customers matter here. Dow's Long Mott project in Texas wants the Xe-100 for industrial steam and power. Amazon invested in X-energy and, in October 2025, described plans with Energy Northwest for the Cascade Advanced Energy Facility near Richland, Washington: an initial four Xe-100 units at 320 MWe, expandable to 12 units and 960 MWe, as part of a goal of more than 5 GW of X-energy capacity by 2039.

Oklo

Oklo submitted the first advanced reactor combined license application, and the NRC denied it in January 2022 for insufficient information, saying the company could reapply. The design has since grown from a 1.5 MWe microreactor to a 75 MWe sodium fast reactor modeled on Idaho's EBR-II. Oklo is now building at INL under DOE authority, which is faster but covers a demonstration. Selling commercial power at scale still runs through the NRC. Sam Altman chaired Oklo's board until April 2025.

Kairos Power

Kairos is building in steps: the Hermes test reactor, then Hermes 2, then commercial plants. Google signed in October 2024 to buy up to 500 MW from Kairos reactors, the first by 2030 and the rest through 2035. A TVA and Google arrangement later raised Hermes 2's planned output to 50 MWe.

The Idaho criticality race

Executive Order 14301 (May 2025) told DOE to get at least three test reactors critical by July 4, 2026. ANS reported that Antares (June 4, at INL), Valar Atomics (June 18, at a Utah site) and Deployable Energy (announced July 1, at INL) reached zero-power criticality by the deadline, and Neutron Bytes reported that Aalo Atomics followed at INL on July 4. These were zero-power tests: they proved the chain reaction works, not that the units can make electricity.

Tech company deals and the Crane restart

Data center demand turned nuclear plants into prized assets. The deals come in very different strengths.

One more point that matters locally: a data center buying power from an existing reactor shifts electricity from one set of customers to another. If a data center contracts the output of a plant that was already running, other customers have to get that power somewhere else. Restarts like Crane and uprates do add supply.

Crane is not the only restart, and the other one shows how slow the last steps can be. Holtec's Palisades plant in Michigan, closed in 2022, won NRC approvals in 2025 to return to operating status. Holtec had aimed to reconnect to the grid by the end of February 2026. Fuel loading began August 30, 2026, then paused after an incident involving a fuel assembly, and as of early October 2026 the plant had not restarted and Holtec had not set a firm date, per ANS reporting.