BasicsSmall modular reactors spent decades as a PowerPoint slide. Nuclear engineers loved the concept. Utilities were polite about it. Then, in a tight window between 2022 and 2026, five separate problems that had blocked commercial SMRs for a generation got solved, more or less at the same time. None of them made headlines outside the trade press, but together they explain why construction is now underway in Wyoming and Tennessee, and why the industry finally has money behind it instead of just enthusiasm. This is the story of those five breakthroughs, what changed, and who forced the change. ⚛️
Each one is a different kind of engineering. Physics, regulation, chemistry, materials science, and, yes, spreadsheets. Skip any one of them and the whole stack collapses. Here’s how it came together.
Passive safety finally satisfies the regulators
For fifty years, reactor safety meant pumps, backup pumps, and diesel generators to run the backup pumps. NuScale Power’s VOYGR design broke that pattern by using convection and gravity instead of machinery. Each 77 MWe NuScale Power Module sits in a pool of water and cools itself through natural circulation, no operator action, no electricity, and no external water source required for at least 72 hours if things go wrong. It’s the kind of passive safety case regulators had never actually had to approve before. 🛡️⚙️
The US Nuclear Regulatory Commission certified the VOYGR design in 2022, making it the first SMR ever approved in the United States, and NRC staff cleared the larger 77 MWe module configuration in May 2025. That approval mattered because it proved a regulator built around large pressurized water reactors could actually bless a design that throws out the entire active-safety playbook. 💧
What passive safety actually removes from the plant:
Emergency diesel generators and their fuel storage requirements
Multiple redundant pump trains and the maintenance burden that comes with them
Reliance on offsite power or operator intervention during the first three days of an event
A large chunk of the containment and control-room complexity that drives up large-reactor costs
Have you noticed how much of the anti-nuclear argument from the 1980s rested on “what if the pumps fail”? That objection gets a lot quieter when there aren’t any pumps to fail.
A licensing rulebook built for reactors that aren’t giant
Regulation was arguably the bigger bottleneck than physics. 🏛️ The NRC’s rulebook was written for gigawatt-scale light-water plants, and every SMR developer had to argue, case by case, why their much smaller, differently cooled reactor deserved different treatment. That changed in March 2026, when the NRC finalized 10 CFR Part 53, a technology-inclusive, risk-informed framework built specifically for advanced reactors rather than shoehorned around the old one. The rule arrived more than a year ahead of the 2027 deadline Congress set in the 2019 Nuclear Energy Innovation and Modernization Act, and it followed the 2024 ADVANCE Act, which cut advanced-reactor fees and pushed the agency to move faster. 📜⚖️
The proof that the new posture works showed up in the construction permits, not just the paperwork. Kairos Power received a construction permit for its 35-MWt Hermes demonstration reactor in Oak Ridge, Tennessee, back in December 2023, the first non-light-water reactor construction approval in the US in decades. Then in March 2026 the NRC approved a construction permit for TerraPower’s Natrium plant in Kemmerer, Wyoming, its first approval for any commercial non-light-water reactor design in more than 40 years, and its first commercial reactor approval of any kind in nearly a decade.
What actually shifted at the agency:
Reduced hourly fees for advanced-reactor applicants and pre-applicants
A technology-inclusive licensing path instead of forcing every design into the light-water mold
Faster review timelines, with TerraPower’s permit review compressed by roughly eight months
New guidance for microreactors and for siting reactors on brownfield and retired fossil-fuel sites
The fuel that took a decade to catch up
You can certify the reactor and streamline the paperwork, and none of it matters if there’s no fuel. Most advanced SMR designs, including Kairos’s Hermes and X-energy’s Xe-100, run on HALEU, uranium enriched between 5% and 20%, well above the roughly 5% used in today’s commercial fleet. For years the only large-scale HALEU producers were in Russia and China, which is an awkward supply chain for a US industry the government is trying to build up. 🔬
Two things broke the logjam. First, the Department of Energy committed $2.7 billion over ten years in January 2026 to expand domestic enrichment capacity, on top of Centrus Energy’s Piketon, Ohio, facility, which had already produced more than 920 kilograms of HALEU by mid-2025. Second, and more concretely, X-energy received a Special Nuclear Material License from the NRC in February 2026 for its TX-1 facility, making it the first-ever Category II nuclear fuel facility licensed in the United States and clearing the way to commercially manufacture TRISO-X fuel, the tristructural-isotropic pebble design that encapsulates uranium in layers of graphite and silicon carbide. ⚗️
Why the fuel breakthrough matters this much:
HALEU enables higher burnup and smaller cores, which is part of what makes an SMR small
TRISO fuel’s ceramic coating is designed to contain fission products even at extreme temperatures
A domestic Category II license removes a Russia- and China-shaped hole in the supply chain
ASP Isotopes alone reports customer interest topping $30 billion for its laser-based HALEU enrichment process, a sign the demand side isn’t waiting around
If you want the fuller picture of who’s licensed what and when across every HALEU project in motion, SMRbrief Pro members can search, filter, and export the full intelligence picture behind developments like this one, which is more useful than trying to track it across a dozen NRC dockets yourself. 🧪
Coolants beyond water finally leave the lab
Light water is cheap and well understood, but it also caps how hot a reactor can run and forces heavy, high-pressure vessels. The SMR field’s answer has been to try coolants that don’t need pressure at all. Kairos Power’s Hermes reactor uses FLiBe, a molten mixture of lithium and beryllium fluoride salts, to cool a bed of TRISO fuel pebbles near atmospheric pressure, and the company has now completed the first safety-related concrete pour at its Oak Ridge site, anchored on 51 drilled piers sunk roughly 40 feet into bedrock. 🌡️
TerraPower took a different but related path. Its Natrium reactor at Kemmerer, Wyoming, pairs a 345-MWe sodium-cooled fast reactor with an integrated molten salt heat storage system that can push output as high as 500 MWe on demand, letting the plant act almost like a battery plugged into the grid. Construction officially began there on April 23, 2026, with Bechtel as the engineering and construction partner, and TerraPower CEO Chris Levesque called it the moment the industry had been “working toward for a generation.” Over in the UK, Rolls-Royce SMR’s 470 MWe pressurized-water design is working through the Office for Nuclear Regulation’s Generic Design Assessment, with roughly 90% of the plant meant to be built in factory conditions rather than poured on site. Three coolants, three regulators, one very busy decade. ❄️🏗️
A quick comparison of what’s actually under construction or under review right now:
Hermes: 35 MWt, FLiBe molten salt coolant, TRISO pebble fuel, Oak Ridge, Tennessee
Natrium: 345 MWe (boostable to 500 MWe), sodium-cooled fast reactor with molten salt storage, Kemmerer, Wyoming
VOYGR: 77 MWe per module, light-water, passively safe, NRC-certified since 2022
Rolls-Royce SMR: 470 MWe, light-water, in UK regulatory review, roughly 90% factory-built
That’s four genuinely different engineering bets running in parallel, on two continents, with three separate regulators watching. Which of these designs do you think reaches commercial operation first?
Hyperscaler cash de-risks the first-of-a-kind problem
None of the last four breakthroughs pay for themselves. A first-of-a-kind SMR plant can cost $1 billion to $3 billion, and utilities have historically been reluctant to shoulder that risk alone, which is exactly what killed NuScale’s original Idaho project when costs climbed from $3.6 billion to $9.3 billion. What changed the financing math wasn’t a new reactor at all. It was AI. 💰🔌
Data centers now need power at a scale and reliability that public grids increasingly can’t guarantee, so hyperscalers started signing long-term power purchase agreements that function as loan guarantees in disguise. Amazon signed a 1.92 GW agreement with Talen Energy, and Microsoft locked in 837 MW with Constellation Energy, part of a broader trend that has pushed data-center-linked SMR offtake agreements from roughly 25 GW in late 2024 to nearly 45 GW today. A creditworthy tech company promising to buy the power for twenty years turns a speculative reactor into a bankable asset, which is precisely the kind of revenue certainty that project financiers demand before they’ll touch a first-of-a-kind build. 📈
What the hyperscaler money actually unlocks:
Revenue certainty that lets developers secure construction financing without waiting for a utility rate case
A customer willing to pay a premium for round-the-clock, carbon-free power near their own facilities
Enough deal volume that supply chains for TRISO fuel, forgings, and skilled labor can plan around real orders
Political cover, since jobs and AI competitiveness make advanced nuclear funding an easier sell in Congress
Five breakthroughs, one narrow window, and a genuinely uncertain outcome, since cost overruns and construction delays have killed SMR projects before and could again. But for the first time, the physics, the paperwork, the fuel, the hardware, and the money are all pointing the same direction at once. Which of these five pieces would you bet on breaking first if something has to give?