ProjectsBasicsPlant Vogtle cost $36.8 billion to build two reactors. That is the number every SMR pitch deck is quietly reacting to, whether it says so or not. The traditional large-reactor model produced the most expensive power plant in human history, and the industry needed a different story to tell investors, regulators, and the public. Small modular reactors became that story. Whether they can actually deliver on it is a separate question, and a genuinely interesting one. 💡
Here is what is actually driving the “future of nuclear” label, and where the label is still ahead of the concrete. ⚛️
Traditional nuclear broke its own economics
Georgia Power’s Vogtle Units 3 and 4 were approved in 2012 at an estimated $14 billion. By the time Unit 4 entered commercial service in April 2024, the total price tag had climbed past $36.8 billion, more than double the original estimate, and both units arrived more than six years late. Westinghouse, the lead contractor, went bankrupt in 2017 partway through construction. 📉
That is not one bad project. It is the pattern. Big, bespoke, site-built reactors have run over budget and behind schedule on nearly every Western project of the last two decades, and ratepayers in Georgia are still covering the bill through monthly surcharges. SMR developers point to that record constantly, and fairly, because it is the single best argument for trying something structurally different: factory-built modules, standardized designs, and smaller individual price tags that do not sink a utility’s balance sheet if one project runs long. 🏗️
Vogtle Units 3 and 4: $14 billion approved cost in 2012, $36.8 billion actual final cost
Timeline: originally targeted for 2016 and 2017, actually completed in 2023 and 2024
Contractor risk: Westinghouse bankruptcy in 2017 mid-construction
The pitch: smaller, factory-built SMR units are supposed to cap that kind of downside risk per project
Worth sitting with: the case for SMRs is not really that nuclear got cheaper. It is that the old way of building nuclear became too financially dangerous for most utilities to attempt again.
AI data centers gave SMRs a customer with real money
Big Tech is not funding SMRs out of climate conscience alone. It is funding them because AI data centers need firm, round-the-clock power that the grid cannot supply fast enough. Across 13 tracked deals, the four major US hyperscalers have committed more than 9.8 GW of nuclear capacity to AI infrastructure as of May 2026, according to SMR Intel’s deal tracker. ⚡
Google signed the first corporate agreement to buy power from multiple SMRs, a Master Plant Development Agreement with Kairos Power for 500 MW of capacity by 2035, with the first reactor targeted for 2030. Amazon led a $700 million investment in X-energy for up to 12 Xe-100 reactors and is developing a four-unit, 320 MW project with Energy Northwest near Richland, Washington. Meta holds the largest single commitment by volume, up to 6.6 GW across agreements with TerraPower, Oklo, Vistra, and Constellation, though delivery stretches out to 2032 through 2035. 🚀
Total hyperscaler nuclear commitment: 9.8 GW across 13 deals tracked through May 2026
Google and Kairos Power: 500 MW KP-FHR agreement, first reactor targeted 2030
Amazon and X-energy: $700 million investment, up to 12 SMRs planned
Meta: up to 6.6 GW across four separate developers, the largest volume commitment of any hyperscaler
That demand is why the global SMR project pipeline reached 47 GW by the end of the first quarter of 2026, per Wood Mackenzie, with more than half of it sitting in the United States. A technology needs paying customers to become an industry, not just a good idea, and AI just handed SMRs their first real ones.
Regulators built a shortcut that only SMRs could use
Part 50 and Part 52, the NRC’s existing licensing frameworks, were written for large light-water reactors in 1956 and 1989. Every advanced design that did not fit that mold had to negotiate exemptions case by case, which is exactly why TerraPower’s Natrium reactor took years of pre-application work before it could even file for a construction permit. 🏛️
That changed on April 29, 2026, when the NRC’s new 10 CFR Part 53 framework took effect, more than a year ahead of the deadline Congress set in 2019. Part 53 is risk-informed, performance-based, and technology-inclusive by design, meaning a molten salt reactor or a sodium-cooled fast reactor no longer has to argue its way around rules built for pressurized water plants. It is optional. Developers can still use Parts 50 or 52 if that fits their design better. But for the first time, US regulators built a path with non-light-water SMRs specifically in mind, rather than forcing them to fit an old one. 🔬
Part 50 and Part 52: original licensing frameworks, dating to 1956 and 1989, built around large LWRs
Part 53: new optional framework, effective April 29, 2026, risk-informed and technology-inclusive
Legislative origin: mandated by the 2019 Nuclear Energy Innovation and Modernization Act, reinforced by the 2024 ADVANCE Act
Effect: non-light-water SMR designs can now be licensed without the exemption-heavy process older frameworks required
Britain and Canada are running their own versions of the same shift. The UK’s Generic Design Assessment and Canada’s Canadian Nuclear Safety Commission have both moved SMR-specific projects through construction licensing faster than any large reactor has managed there in decades, which tells you the regulatory bottleneck was never really about safety physics. It was about paperwork built for a different machine.
The buildout is already global, not theoretical
This is not one country making a bet. Russia’s Akademik Lomonosov has supplied power to the Arctic town of Pevek since 2020, still the only operating commercial SMR anywhere. China’s Linglong One, a 125 MWe reactor at the Changjiang site in Hainan Province, is targeting commercial operation in the first half of 2026, which would make it the world’s first land-based commercial SMR. Canada’s Ontario Power Generation is pouring concrete on a BWRX-300 unit at Darlington, the first new nuclear construction in Ontario in more than three decades. 🌍
The United Kingdom signed a firm contract in April 2026 to build three Rolls-Royce SMR units at Wylfa, backed by £2.6 billion in government funding. The United States holds the first NRC-certified SMR design, NuScale’s VOYGR, and broke ground on TerraPower’s Natrium reactor in Wyoming the same month the UK signed its Wylfa deal. Five countries, five different reactor technologies, all moving at the same time.
Russia: only country with an operating commercial SMR since 2020
China: closest to the next commercial milestone, targeting H1 2026
Canada: first G7 nation with SMR construction actively underway
UK and US: leading on funding commitments and regulatory certification, respectively
The label is still ahead of the delivery record
Here is the honest complication nobody selling an SMR wants dwelt on for too long: not one Western SMR design has actually generated commercial power yet. NuScale’s original Idaho project, the Carbon Free Power Project, collapsed in November 2023 after its target price climbed from $58 to $89 per megawatt-hour, and every cost estimate published since carries a first-of-a-kind premium that ranges from $80 to $180 per megawatt-hour, according to multiple 2026 techno-economic analyses. That is not cheap yet. It might not be cheap for a while.
The case for calling SMRs the future of nuclear rests on trajectory, not current pricing: shorter build times, smaller absolute financial exposure per project, a regulatory framework finally built for the technology, and a paying customer base that did not exist five years ago. Whether that trajectory holds depends entirely on whether Kemmerer, Darlington, and Wylfa actually finish on budget, something no Western SMR project has proven yet at scale. If you’re tracking the nuclear market professionally, SMRbrief Pro gives you the structured database to go deeper than any single article can, especially as these first-of-a-kind projects start reporting real construction numbers instead of projected ones.
So which milestone actually earns the label: the day a Western SMR connects to the grid, or the day one of them does it on budget?