🏛️ Issues and Advocacy Training
Courses › Supplemental Nuclear: Small Modular Reactors

Lesson 4 of 9 · about 18 minutes

Projects planned, deals signed and the cost record

You'll be able to

  • List the main North American SMR projects and their status as of October 2026
  • Sort a nuclear deal by firmness, from construction under way to memorandum
  • Compute cost per kilowatt from Darlington's published numbers and explain why later units should cost less
  • State the supporters' and critics' strongest arguments and where the evidence points

After years of announcements, 2026 was the year concrete started showing up. Ontario lifted a reactor basemat into its shaft in April. TerraPower started nuclear construction in Wyoming the same month. The NRC issued TVA a permit in September. And still, as of October 2026, no SMR anywhere in North America was sending power to a grid. This lesson lists the projects, sorts the deals and looks at the first real cost numbers.

The project list

ProjectDesignOwner or buyerStatus, October 2026
Darlington New Nuclear, Ontario4 BWRX-300, about 1,200 MWe totalOntario Power GenerationUnder construction since 2025. Basemat lifted in April 2026; OPG's goal is grid connection of the first unit by the end of 2030.
Clinch River, TennesseeBWRX-300TVANRC construction permit, September 2026. DOE selected TVA for up to $400 million in cost-shared funding in December 2025. No construction date; TVA negotiating cost-sharing partners.
Kemmerer Unit 1, WyomingNatrium, 345 MWeTerraPower subsidiary, with PacifiCorp's retiring coal site nearbyConstruction began April 23, 2026; completion expected 2030.
Long Mott, Seadrift, Texas4 Xe-100, 320 MWeDow (Long Mott Energy)NRC permit review; final safety evaluation targeted November 2026.
Hermes 2, Oak Ridge, TennesseeKairos, aimed at 50 MWeKairos, with Google and TVA tied to outputGroundbreaking April 2026; operations expected by 2030.
Palisades SMR-300s, Michigan2 Holtec SMR-300, about 340 MWe eachHoltecPartial construction permit application accepted early 2026; NRC final reviews expected March 2027.
Cascade, near Richland, WashingtonXe-100, initial 320 MWe, up to 960Energy Northwest, Amazon-backedConstruction permit application expected first half of 2027, per X-energy in June 2026.
Aurora-INL, IdahoOklo, up to 75 MWeOkloUnder construction under DOE authorization; no operating date.
Army Janus programMicroreactors from five vendorsU.S. Army, up to $2.2 billion through fiscal 2031Vendors and bases named August 26, 2026; first service targeted September 2028.
Carbon Free Power Project, INL6 NuScale modules, 462 MWeUtah Associated Municipal Power SystemsCancelled November 2023 over cost.

Data center and corporate deals, sorted by firmness

The earlier SMR lesson walked through Meta's "up to 6.6 GW" headline. The same method works for any deal: put each piece in one of these bins.

  1. Construction under way, with a regulator's permit. Darlington; Kemmerer.
  2. Permit in hand or near, money committed, construction not started. Clinch River; Long Mott.
  3. Signed purchase agreement for reactors not yet permitted. Google's October 2024 agreement to buy up to 500 MW from Kairos reactors through 2035.
  4. Equity investment plus a planned project. Amazon's stake in X-energy and the Cascade plan.
  5. Framework agreements and rights to future units. TerraPower and Meta's January 9, 2026 agreement covering eight Natrium plants, two funded and energy rights to six more; Meta's planned Oklo campus in Ohio.
  6. Stated plans and memorandums. Aalo's plan for Aalo-X to power a co-located data center in 2027; many announced partnerships with no money or permit attached.

Most of the gigawatts in the headlines live in bins 4 through 6. They may still get built, but they depend on licensing, fuel and financing that haven't happened yet.

The first real cost numbers

Before 2025, SMR costs were vendor projections. Now there are two public data points worth knowing: the NuScale project that died and the Darlington project being built.

Example: Darlington per kilowatt

OPG lists C$6.1 billion for the first BWRX-300 plus C$1.6 billion for systems shared by all four, and C$20.9 billion for all four.

  • First unit with common costs: C$7,700,000,000 ÷ 300,000 kW ≈ C$25,700 per kW.
  • First unit alone: C$6,100,000,000 ÷ 300,000 kW ≈ C$20,300 per kW.
  • All four: C$20,900,000,000 ÷ 1,200,000 kW ≈ C$17,400 per kW.
  • Implied cost of units 2 to 4: C$20.9 billion minus C$7.7 billion = C$13.2 billion, or C$4.4 billion each, about C$14,700 per kW.

So OPG expects each later unit to cost roughly 28 percent less than the first one alone (C$4.4 billion versus C$6.1 billion). That is the learning curve SMR supporters predict, written into a real budget. Whether it holds is what the next four years will show. For scale, the costs and timelines lesson puts Vogtle 3 and 4 at about US$15,700 per kW. These are Canadian dollars, so convert before comparing.

Example: the NuScale lesson in one division

The Carbon Free Power Project's estimate rose to $9.3 billion for 462 MWe. $9,300,000,000 ÷ 462,000 kW ≈ $20,100 per kW. The target power price rose from $58 to $89 per MWh, and the $89 already counted an estimated $30 per MWh subsidy from the 2022 Inflation Reduction Act, so the unsubsidized figure was near $119. UAMPS member utilities had the right to withdraw, enough did, and the project ended in November 2023. POWER Magazine reported the project had received $232 million in DOE support by then. The NRC never rejected anything. The buyers walked.

The NuScale case matters in Idaho because the plant was to be built at Idaho National Laboratory, and the participants were public power systems in the region. The lesson local officials drew: public power members carry real risk until they have a firm price and a firm exit date.

The strongest case on each side

Supporters argueCritics argue
Nuclear runs around the clock; the U.S. fleet averaged a 91 percent capacity factor in 2025, per EIA. Data centers and factories need that.Reliability is wasted if the plant arrives in 2035 at twice the price. Gas, batteries and grid upgrades can be built in a few years.
Carbon-free power with a small land footprint.Cheaper carbon-free options exist now; every dollar spent on a late reactor is a dollar not spent on them.
Permanent, well-paid operating jobs, often at retiring coal sites like Kemmerer, reusing grid connections and workers.Construction jobs are temporary, and operating staff per megawatt is high for small units, which hurts cost.
Smaller projects put less money at risk per project, and repeat orders cut costs (OPG's own budget shows it).First units have come in high per kilowatt (NuScale, Darlington), and the 2020 MIT study of U.S. history found little learning between plants.
Tech companies can absorb first-of-a-kind premiums so ratepayers don't.Deals are often options and memorandums; if a buyer walks, regulated customers or taxpayers can be left holding costs.
Passive safety and smaller cores justify smaller emergency zones (lesson 6).Waste per unit of energy may be higher for some SMRs (lesson 9), and there is still no repository.

The supporters' side is carried by the Nuclear Energy Institute, utilities like TVA and OPG, and DOE, whose Liftoff report argues a committed order book of five to ten deployments of one design is what drives costs down. The skeptics' side is carried by the Institute for Energy Economics and Financial Analysis, which tracked the NuScale cost increases, and the Union of Concerned Scientists.

Where the evidence points: on reliability and carbon, the supporters are right about the existing fleet. On cost, first units in North America have been expensive, and the cost declines are a forecast that only a built series can prove. On timelines, the 2026 permits and construction starts are real progress, and every commercial SMR still sits at least four years from power. The honest position for a local official is that SMRs are worth watching and worth planning for, and that no community should take on cost risk for a first-of-a-kind unit without a firm price, a firm exit and someone else holding the overrun.

Common mix-ups

  • Claim: NuScale failed because the NRC blocked it. What the evidence shows: The NRC approved the design. The project ended when the price rose and buyers withdrew.
  • Claim: Data center companies have ordered dozens of SMRs. What the evidence shows: Most deals are options, equity stakes or framework agreements. Few are firm orders with construction under way.
  • Claim: SMRs are already producing power in North America. What the evidence shows: As of October 2026 none was. Darlington's first unit targets grid connection by the end of 2030.

Reporting angle

  • For any announced deal, ask which bin it belongs in: permit, contract, equity, option or memorandum. Ask for the document.
  • Ask the developer for cost per kilowatt in its own numbers, in what year's dollars, and whether common and financing costs are included.
  • For public power or co-op participation, ask about withdrawal rights, the deadline to exercise them and what members owe if they leave.
  • At the state utility commission, watch for integrated resource plans or certificate filings that list an SMR. That is where costs enter regulated rates.
  • Ask local proponents to explain the NuScale cancellation and what would be different this time. A team that can't answer that hasn't done its homework.

Research angle

Check yourself

1. OPG's published cost for the first Darlington BWRX-300 is C$6.1 billion plus C$1.6 billion in common costs. About how much is that per kilowatt for a 300 MWe unit?

2. Why was the NuScale Carbon Free Power Project, planned at Idaho National Laboratory, cancelled in November 2023?

3. Which project had nuclear construction under way in the United States in October 2026?

4. OPG's estimate for four BWRX-300s is C$20.9 billion. What is the average per kilowatt across all four?

5. TVA's stated priority for Clinch River financing, per World Nuclear News, is what?

6. Where does the evidence point most clearly on SMR costs as of October 2026?

Terms in this lesson

First-of-a-kind (FOAK)
The first units of a new design, which carry design completion, supply chain setup and learning costs.
Common costs
Site-wide costs such as intake structures, roads and support buildings shared by several units.
Subscription
In public power projects, the share of a plant's output that member utilities agree to buy.
Memorandum of understanding (MOU)
A non-binding statement of intent to work together. Not a purchase, order or contract.
Other Transaction Authority (OTA) agreement
A flexible federal contract form used by DOE and the military for prototypes, outside standard procurement rules.

Full glossary

Sources