
Photo: U.S. Nuclear Regulatory Commission, 2023, via Wikimedia Commons (CC BY 2.0)
At a site near Kemmerer, Wyoming, beside a retiring coal plant, TerraPower is building what it says will be the first utility-scale advanced nuclear power plant in the United States. The reactor, called Natrium, is cooled by liquid sodium instead of water and is paired with a tank of molten salt that works like a heat battery. The U.S. Nuclear Regulatory Commission gave it a construction permit in March 2026, and the project has had a busy autumn: its main contractor stepped away from the next phase, a new regulatory finding came in, and a fuel supply that does not fully exist yet stayed at the center of the story.
The short answer: The Nuclear Regulatory Commission voted on March 4, 2026 to approve the construction permit for Kemmerer Unit 1 and issued it on March 9. TerraPower says construction of the reactor plant began on April 23. The unit is designed to deliver 345 megawatts of steady electricity and to boost output to 500 megawatts for more than five hours using stored heat. The company’s public target is to finish in 2030, with commercial operation in 2030 to 2031. The plant is not running and has no operating license yet. TerraPower expects to apply for that license in 2028. The project’s estimated cost is about $4 billion, and the U.S. Department of Energy is paying up to about $2 billion of it. The main open risks are the construction contract, which TerraPower is rebidding after Bechtel stepped away in September, and the supply of the special fuel the reactor needs.
Below: how the plant works, a dated timeline of how it got here, answers to the questions people ask most, and what is still unresolved.
How Natrium works: a fast reactor with a salt battery
Natrium has two halves that are built to be separate. The nuclear half is a sodium-cooled fast reactor. The other half is an energy island with a steam turbine and two large tanks of molten salt. Heat from the reactor can either go straight to the turbine or be parked in the salt and used later.
The reactor: liquid metal instead of water
In a typical commercial reactor, the water coolant has to be pressurized to roughly 100 to 150 times normal atmospheric pressure so it does not boil away, according to Argonne National Laboratory. A sodium fast reactor does it differently. Sodium is a metal that is liquid at reactor temperatures, it carries heat very well, and its boiling point is far above the temperature the core runs at. Argonne puts the normal coolant temperature around 500 degrees Celsius (930 degrees Fahrenheit) and sodium’s boiling point near 900 degrees Celsius (1,650 degrees Fahrenheit), so the coolant can stay at normal atmospheric pressure.
TerraPower describes Natrium as a pool-type reactor, meaning the core sits submerged in a tank of sodium with no pipe penetrations through the vessel below the lid. The company says this removes the possibility of a leak that drains the coolant, and that the design relies on gravity and rising hot air to remove heat after a shutdown without needing electric power. Those are TerraPower’s design claims. The NRC reviewed the design before issuing the construction permit, but no Natrium reactor has operated yet.
The storage: heat held in molten salt
The second half is what makes Natrium unusual. The reactor can run at a steady output while a molten salt system takes some of its heat. When the grid needs more power, for example in the evening after solar output fades, the plant sends more heat from the salt to the turbine. TerraPower says that lifts output from 345 megawatts to 500 megawatts for more than five hours. The salt is held in its own system, separated from the sodium and the nuclear island, which also lets the turbine side be run outside the nuclear control area.
TerraPower argues this is meant to fit a grid with a lot of wind and solar. That is the company’s pitch, and it is not yet proven at a real plant. Whether it works as advertised is one of the things Kemmerer is there to demonstrate.
The fuel: HALEU metal
Natrium burns high-assay low-enriched uranium, usually shortened to HALEU. TerraPower defines it as uranium with more than 5 percent but less than 20 percent of the fissile isotope uranium-235, compared with the roughly 5 percent or less used in today’s large reactors. The fuel is a metal alloy rather than the ceramic pellets used in most conventional reactors. This is the most important practical constraint on the whole project, and it comes up again below.
Why it matters that sodium reactors have been done before
Natrium is new as a commercial product but it is not a new kind of physics. Its lineage runs back to sodium-cooled research reactors such as Experimental Breeder Reactor II, a 20-megawatt sodium-cooled research reactor at Argonne’s Idaho site that ran from 1964 until it was shut down in 1994.

Photo: U.S. Department of Energy, via Wikimedia Commons (Public domain)
In 1986, Argonne ran two full-power tests on EBR-II with its automatic shutdown systems deliberately disabled. In one, the team cut power to the pumps that push coolant through the core. In the other, they cut the secondary coolant pump so heat was no longer being removed from the primary system. Argonne reports that in both cases the temperature rose briefly, then the reactor’s own feedback effects took over and it cooled, and that it shut itself down without operators or emergency systems.
Back in 1986, we actually gave a small prototype advanced fast reactor a couple of chances to melt down. It politely refused both times. (Pete Planchon, Argonne nuclear engineer who led the 1986 tests, Argonne Logos, Winter 2002)
Those tests are the best argument for the sodium approach, and they came from a small reactor with metal fuel, the same combination Natrium uses. They are also from 1986 and from a 20-megawatt machine, not a 345-megawatt commercial plant.
Sodium has a downside that the same Argonne article is candid about: it can be dangerous if it touches air or water. The clearest warning is Japan’s Monju prototype reactor. On Dec. 8, 1995, a broken thermometer sheath in a secondary sodium loop leaked about 640 kilograms of hot sodium. The sodium caught fire on contact with air. A case study by Hiroshi Iino of the Kanazawa Institute of Technology says the delay in shutting the reactor down let the leak run for about three hours, and that air conditioning left running kept supplying fresh air to the fire. The sodium was not radioactive, but the fire damaged the surrounding structure, and the operator’s handling of information afterward, including edited video and late notice to neighbors, set off public protest. That is the kind of ordinary engineering and operations problem that a first-of-a-kind sodium plant has to avoid.
The Natrium timeline: what has happened and when
September to October 2020. TerraPower and GE Hitachi unveil Natrium, and the Department of Energy picks it as one of two flagship projects under its Advanced Reactor Demonstration Program, with roughly $80 million in initial funding.
November 2021. TerraPower and PacifiCorp name Kemmerer, Wyoming, near the Naughton coal plant, as the preferred site.
March 2024. TerraPower submits its construction permit application. The NRC’s project page lists it as tendered on March 28, 2024 and docketed on May 21, 2024. The non-nuclear groundbreaking at the site came in June 2024.
October to December 2025. The NRC finishes its final environmental impact statement in October and its final safety evaluation in December, according to the agency’s schedule.
January 9, 2026. Meta and TerraPower announce an agreement for up to eight Natrium plants, covered below.
March 4 and 9, 2026. The commissioners vote to approve the permit, and the NRC issues it five days later. The Department of Energy calls it the first construction permit the NRC has issued for a commercial non-light-water power reactor. The review took about 18 months against an original schedule of 27, per TerraPower.
April 23, 2026. TerraPower announces the official start of construction on Unit 1. POWER magazine reports the work is expected to bring about 1,600 workers to the site during construction and about 250 permanent staff once the plant operates.
August 2026. TerraPower names 12 equipment vendors for the Natrium plant and the sodium facility. Separately, Hyundai Engineering and Construction is selected as the construction contractor for up to eight future Natrium reactors, a framework with completion, price and performance guarantees that TerraPower says are meant to support conventional financing.
September 22, 2026. TerraPower says it will rebid the construction contract for Kemmerer after it and Bechtel decided not to go into the full construction phase together.
Early November 2026. Curtiss-Wright reports it has finished installing a full-scope training simulator at the Kemmerer site. Then on November 3, the NRC staff accepts, with five conditions, a TerraPower report on a two-unit Natrium layout.
Questions people ask about the Kemmerer reactor
Is Natrium running yet?
No. Kemmerer Unit 1 is a construction site, not a working plant. Having a construction permit means the NRC has agreed the plant can be built; it does not allow it to operate. TerraPower will need to apply for a separate operating license, which the company says it expects to do in 2028. Restarts of existing plants, like the Three Mile Island restart, are a different and faster path. Another new design, the Tennessee Valley Authority small modular reactor, has also so far reached only the construction permit stage.
When will it make power?
TerraPower’s public target is completion in 2030, and it has pointed to commercial operation in 2030 to 2031. POWER magazine reports that the plant’s owner told the NRC it expects to finish construction by February 2031. Those are all targets, and this is a first-of-a-kind plant. The project’s own first date slipped before: a 2023 announcement moved the completion target from 2028 to 2030, which POWER attributes mainly to concerns about fuel availability.
How much will it cost?
ENR puts the estimate at about $4 billion. Under the Department of Energy’s demonstration program, up to roughly $2 billion is federal money, matched dollar for dollar by TerraPower and its partners. TerraPower says the first-of-a-kind cost includes work that later plants will not repeat, such as design and licensing, building a sodium test facility and developing fuel fabrication. That is why the number is not a good guide to the price of the second plant.
What happened with Bechtel?
Bechtel had worked with TerraPower since 2020 and was the construction contractor of record for the early phases. On September 22, TerraPower said it would rebid the full construction contract and that Bechtel would stay on to finish the Sodium Test and Fill Facility. Bechtel did not give a reason, and TerraPower did not in the statements we saw. Bechtel’s statement said the work to date was done under an earlier phase of the contract, and that "Bechtel and TerraPower decided to move forward separately."
A Bechtel notice filed in Virginia listed about 200 positions as potentially affected starting around November 16. Bechtel now says those employees will move to other nuclear projects and that it does not expect actual layoffs. TerraPower says construction will keep its current schedule with the subcontractors already on site, and it has not named who will bid or when the contract will be awarded.
What is Meta’s deal?
On January 9, Meta said it would fund early development of two new Natrium units that could provide up to 690 megawatts of firm power as early as 2032, and that it would get rights to energy from up to six more units, 2.1 gigawatts, targeted for 2035. Across all eight possible units, the companies describe 2.8 gigawatts of baseload power plus extra storage capacity that could lift peak output to 4 gigawatts. Nothing in the agreement is a finished plant. The companies said they would choose a site for the first pair in the coming months, and we did not find an announced site in the sources we checked.
What did the NRC decide on November 3?
According to reports from Nuclear News Network and CiVL, the NRC staff found a TerraPower report titled the Natrium Generic Dual Unit Topical Report acceptable for future licensing applications to reference, subject to five conditions. The design in the report has two reactor units that share a fuel-handling building and a common main control room. The conditions require future applicants to show how they meet the design criteria, document the control room’s human factors, analyze shared electrical systems, show that an accident in one unit cannot stop the other from reaching safe shutdown, and address any links between the units’ salt tanks. This does not approve any plant for construction or operation. It gives later applicants, including a possible Meta site, a defined starting point.
Where does the fuel come from?
HALEU is the weak link in the whole American advanced reactor effort, and TerraPower has tried to line up several suppliers at once. The plan has the uranium enriched, turned into metal at a Framatome plant in Richland, Washington, and then fabricated into fuel assemblies at the Natrium Fuel Facility being built at Global Nuclear Fuel’s site in Wilmington, North Carolina. Framatome and TerraPower reported in November 2025 that they had produced test uranium metal pucks on the Richland line, using depleted uranium rather than HALEU.
For the enrichment step, ASP Isotopes announced agreements with TerraPower in May 2025: one supply agreement for the first fuel core, and a 10-year agreement for up to 150 metric tons of HALEU from 2028 through 2037. ASP’s planned plant at Pelindaba, South Africa, is meant to produce about 15 metric tons a year and to start in 2027, but the company itself says that depends on receiving all required permits and licenses. ASP described itself as a development-stage company when it announced the deal, and the Pelindaba plant was still to be built. TerraPower has also worked with Centrus, whose planned new enrichment capacity in Ohio was reported to be expected in 2029.
Does it replace wind and solar?
No, and TerraPower does not say it does. Its argument is that a plant which can hold steady and then surge for hours is a complement to variable renewables on a grid that also has to serve data centers and electrified demand. The company’s FAQ describes the energy storage as holding much more energy than typical battery installations. Other long-duration storage approaches are competing for the same job, including the iron-air batteries from Form Energy and the sodium-ion batteries now moving toward grid use.
Having a construction permit means the NRC has agreed the plant can be built; it does not allow it to operate.
What is still unresolved
The permit is real and the construction is real, but much of what makes Natrium valuable is still on paper. Four open items matter most:
The construction contract. The full contract for Kemmerer Unit 1 is out for rebid, with no named bidders and no award date. TerraPower says the schedule is unaffected, but a new lead contractor takes time to ramp up.
The fuel. The first core needs HALEU metal on a schedule that depends on enrichment plants that are still planned. No one outside the companies can yet say the 2030 date is safe from that.
The operating license. TerraPower expects to file in 2028 and the NRC has not yet reviewed it. The permit’s review was ahead of schedule, which is encouraging, but a different, longer review of the finished plant is still ahead.
The operating record. No Natrium reactor has yet run, and the sodium and salt systems have not yet been operated together at commercial scale. EBR-II is a good precedent, and Monju is a good warning.
As of November 17, 2026, Kemmerer Unit 1 is permitted and under construction, with nameable risks. The near-term markers are more concrete than the 2030 date: who wins the construction rebid, whether the first core’s fuel supply is confirmed, and when TerraPower files for the operating license.
Further reading

Plentiful Energy: The Story of the Integral Fast Reactor
The history of the Integral Fast Reactor program, the Argonne research line that grew out of EBR-II, written for non-specialists.

Atomic Accidents: A History of Nuclear Meltdowns and Disasters
A readable history of what has gone wrong in nuclear engineering, including accidents at early experimental reactors, and what engineers learned from each.

Kasa Smart Plug Mini with Energy Monitoring (4-Pack)
A smart plug that reports how much electricity an appliance uses, a simple way to see what your own household demand looks like.

Tapo Smart Plug with Energy Monitoring, P115 (4-Pack)
Another energy-monitoring plug option for tracking the daily and monthly use of space heaters, computers and other plugged-in loads.

Amazon Smart Thermostat with C-Wire Adapter
A programmable Wi-Fi thermostat that can shift when your heating and cooling run, one practical way to trim household electricity and gas use.
