BasicsAsk five people how long it takes to build a small modular reactor and you’ll get five different answers, mostly because they’re each describing a different clock. There’s the licensing clock, which starts the moment a developer files paperwork with a regulator. There’s the construction clock, which starts when excavators show up. And there’s the clock most journalists actually mean when they say “build,” which runs from the day a company picks a site to the day the reactor delivers its first megawatt to the grid. Conflate those three and you get the wildly different numbers floating around right now, everything from “18 months” to “a decade.” 🕐
The good news is 2026 finally gave the industry enough real data points to stop guessing. TerraPower has a construction permit in hand. GE Vernova Hitachi has steel going into the ground at Darlington. Kairos Power just had to ask regulators for more time. Here’s what each of those actually tells you about how long this takes.
The five phases every SMR timeline goes through
Before comparing projects, it helps to know what’s actually being measured. Every SMR, regardless of reactor type or country, moves through the same broad sequence, though the length of each phase varies enormously depending on whether the design is first-of-a-kind (FOAK) or has already been proven elsewhere. ⚙️
Pre-application engagement: developers meet informally with regulators for years before filing anything, ironing out technical questions early to avoid surprises later
Licensing review: the formal construction permit or combined license application, reviewed against a published (but frequently missed) timeline
Site preparation and construction: excavation, foundations, module fabrication, and the actual assembly of the reactor building
Testing and commissioning: cold functional testing, then hot functional testing, then fuel load, each requiring separate regulatory sign-off
Commercial operation: the date a plant actually starts delivering power under contract, not just generating heat in a test
The phase everyone underestimates is the first one. Companies like TerraPower and X-energy spent years in pre-application meetings with the NRC before ever submitting a formal permit request, and that groundwork is a big part of why their reviews moved faster than the agency’s own generic schedules predicted.
Natrium: two years from NRC docket to broken ground
TerraPower’s timeline is the cleanest data point the industry has right now, because it’s the only commercial advanced reactor to complete the full licensing-to-construction sequence in the U.S. so far. The company submitted its construction permit application for Kemmerer Unit 1 in Wyoming in March 2024. The NRC formally docketed it in May 2024, setting an initial 27-month review schedule. The review actually finished ahead of that mark: the commission approved the permit on March 4, 2026, roughly 22 months after docketing, and 11 percent under budget according to the Department of Energy’s own account. 🏗️
What happened next is the part people miss. TerraPower didn’t wait around. Official construction on the 345 MW sodium-cooled fast reactor began April 23, 2026, less than seven weeks after the permit was issued. A few numbers that anchor this project’s clock:
22 months: NRC review time, from docketing to construction permit
7 weeks: gap between permit approval and construction start
2030: targeted completion for a facility that could cost up to $4 billion
Total elapsed time from Wyoming site selection (November 2021) to targeted completion sits at roughly nine years
That nine-year figure, not the 22-month licensing window, is the number worth remembering if someone asks you how long an SMR “takes.” The permit review is the fast part.
BWRX-300: eight years from reactor selection to first power
GE Vernova Hitachi’s BWRX-300 project at Ontario Power Generation’s Darlington site tells a slower, steadier story, and it’s arguably the more useful comparison because this design is a 300 MWe boiling water reactor built on already-licensed technology, not a brand-new reactor chemistry. OPG selected the BWRX-300 in December 2021. It applied for a construction license in October 2022. The Canadian Nuclear Safety Commission didn’t grant that license until April 4, 2025, roughly 30 months later, following a two-part public hearing process. Construction actually began in May 2025. 🇨🇦
The current projection has the first unit’s construction wrapping up in late 2029, with commercial operation following in 2030, according to GE Vernova. That means:
~30 months: CNSC licensing review, application to license
~4.5 years: construction license to commercial operation
~9 years total: reactor technology selection (2021) to commercial operation (2030)
1.2 GW: total planned capacity once all four Darlington units are running, with units two through four expected online between 2034 and 2036
Notice that even a proven light-water design, backed by decades of BWR operating history, still lands in roughly the same nine-year range as TerraPower’s genuinely novel sodium-cooled reactor. That should tell you something about where the real time sink is in this industry, and it isn’t the reactor physics. It’s regulatory review plus first-of-a-kind construction learning curves.
Why timelines slip: Hermes 1’s 28-month extension
Not every project holds its schedule, and the honest version of this article has to include the one that didn’t. Kairos Power’s Hermes 1, a non-power demonstration reactor at Oak Ridge, Tennessee, received its NRC construction permit in December 2023, the first non-light-water reactor permitted in the U.S. in more than 50 years. The original permit specified a completion deadline of December 31, 2026. Construction broke ground in July 2024, with the company initially targeting operation sometime in 2027. ⏱️
In March 2026, Kairos went back to the NRC and asked for a 28-month extension, pushing the construction deadline to April 30, 2029. The company’s own filing attributed the delay to development challenges tied to the “first-of-a-kind nature of the Hermes Test Reactor facility design and construction,” as World Nuclear News reported. That’s a remarkably candid admission, and it’s worth sitting with:
28 months: the extension Kairos requested, more than doubling the original construction window
50+ years: how long it had been since the U.S. permitted a non-light-water reactor before Hermes, meaning there was no recent domestic playbook to follow
Hermes 1 is a non-power demonstration reactor, meaning even the “easy” precursor to a commercial plant slipped
Hermes 2, the commercial-scale 50 MWe follow-on, is still targeting 2030, a date that now depends on lessons Hermes 1 hasn’t finished teaching yet
This is the pattern anyone tracking SMR schedules needs to internalize: FOAK projects slip because there’s no supply chain memory to draw on yet, not because the underlying physics changed. If you’re trying to model out which projects are likely to hold their dates versus which ones are showing early slip signals, SMRbrief Pro members can search, filter, and export the full intelligence picture behind developments like this one across every active project, not just the three covered here.
Part 53 promises 18-month reviews, but no one’s tested it yet
On March 25, 2026, the NRC finalized 10 CFR Part 53, the first entirely new reactor licensing framework since 1956. It’s technology-neutral, meaning it doesn’t force every design through the same light-water-reactor-shaped review process, and regulators estimate it could cut design review timelines to as little as 18 months while halving application costs, per coverage of the rule. 📜
Nobody has actually run a project through Part 53 start to finish yet, so treat that 18-month figure as an aspiration rather than a track record. What’s true right now, based on the three real projects above:
Licensing review, even on the fast end, runs 22 to 30 months under the current framework
Construction itself, once permitted, adds another 4 to 5 years for FOAK projects
Total site-selection-to-operation timelines cluster around 8 to 9 years for every serious project with a real date attached
Nth-of-a-kind reactors, the second, third, and fourth units built once the supply chain and workforce exist, should run considerably faster, though nobody has one finished yet to prove it
So if you’re asking how long it takes to build an SMR today, the honest answer is somewhere between eight and nine years from a firm site decision to commercial power, with the licensing phase eating roughly two to three of those years and the rest going to construction, testing, and the inevitable FOAK surprises. Which of these three timelines do you think holds, and which one slips next?