You’ve seen the headlines. Google, Microsoft, Amazon—they’re all signing nuclear deals to power their data centres. The pitch is compelling: clean, baseload power running 24/7 from small modular reactors, arriving by 2030.
Except when you actually dig into the numbers, the story changes fast.
This analysis is part of our comprehensive guide on Big Tech’s nuclear power pivot, where we examine the economic realities behind the headlines.
NuScale‘s Idaho project looked promising until costs exploded from $3.6 billion to $9.3 billion and the whole thing got cancelled in 2023. The only NRC-certified SMR design in the United States couldn’t make the economics work, even with over $1.4 billion in government support backing it up.
If you’re planning data centre capacity, that raises a question: do you wait for SMRs sometime in the 2030s, or lock in renewable PPAs you can have today?
This article cuts through the marketing claims. We’re going to show you what SMRs actually cost per megawatt, how long they’ll really take to deploy, and whether the economics make any sense at all. We’ll break down FOAK versus NOAK pricing and work out the opportunity cost of waiting around. If you’re evaluating energy strategy for your infrastructure, you need transparent numbers—not aspirational projections that might never happen.
Let’s get into it.
How much does an SMR cost per megawatt compared to traditional nuclear reactors?
Here’s the uncomfortable truth: SMRs cost significantly more per megawatt than the traditional large nuclear plants they’re supposed to improve upon.
NuScale’s Idaho project escalated to approximately $20,000 per kilowatt before cancellation. Traditional large reactors run $7,675 to $12,500 per kilowatt. That’s roughly three times higher.
The fundamental issue is that smaller reactors sacrifice plant-size economies of scale. A traditional 1,000+ megawatt reactor spreads fixed costs—regulatory compliance, engineering, infrastructure, security—across significantly more capacity. SMR technologies, which are typically 50 to 300 megawatts each, lack this advantage. You’re paying for nuclear regulation complexity multiple times over for the same total output.
Industry advocates argue that factory fabrication will make up for this through “economies of series production.” Here’s the catch: this requires building hundreds of identical units. No nation has achieved this for SMRs yet. NuScale was originally estimated at $4,200 to $5,000 per kilowatt. Instead, costs more than tripled as first-of-a-kind engineering challenges showed up.
As one Lux Research analyst put it: “Cheap nuclear just isn’t in the cards in the next two decades”.
What is the levelised cost of electricity (LCOE) for SMRs and how does it compare to renewable PPAs?
Current SMR LCOE sits around $89 to $102 per megawatt-hour.
Meanwhile, renewable PPAs today offer solar at $30 to $50 per megawatt-hour and wind at $25 to $45 per megawatt-hour.
That’s a striking gap—SMRs cost roughly double to triple what renewables charge, and they won’t arrive until the 2030s whilst renewable capacity is available right now.
NuScale’s expected LCOE started at $58 per megawatt-hour. By July 2021, that rose to $89 per megawatt-hour after accounting for subsidies. Without subsidies, you’re looking at $119 per megawatt-hour. Even at $89 per megawatt-hour with government subsidies, they couldn’t attract enough utility customers. So the project was cancelled.
SMR advocates point out that direct comparison misses capacity factor. SMRs deliver 24/7 baseload power with capacity factors exceeding 95 per cent. Solar and wind are intermittent, with capacity factors of 25 to 40 per cent. For data centres needing continuous power, you need energy storage. Battery systems cost approximately $115 per kilowatt-hour.
Even accounting for storage costs, renewable combinations often remain competitive with FOAK SMR projections. And that’s assuming SMRs hit their projected LCOE—which NuScale couldn’t.
Translation: SMR economics don’t work without substantial ongoing government subsidies, whilst renewable costs keep declining without them. That’s a significant difference when you’re evaluating long-term cloud computing costs.
What is FOAK financing and why does it matter for SMR deployment costs?
FOAK—First-of-a-Kind—financing addresses the unique problem of funding the first commercial deployment of new nuclear technology. Traditional project finance doesn’t work because lenders won’t accept the combined risks. You need government support.
The U.S. Department of Energy committed over $1.4 billion to NuScale’s Idaho project. Despite this support, the project failed. By cancellation, NuScale had received $232 million—not small change, yet it couldn’t bridge the gap between projected costs and economic viability.
The fundamental issue is that FOAK projects cost 2 to 5 times more than eventual NOAK (Nth-of-a-Kind) pricing. Engineering changes emerge during first construction. Supply chains need development. Regulatory processes involve learning curves. All this translates to cost overruns and delays.
Financing costs make the problem worse. FOAK nuclear projects face interest rates of 8 to 12 per cent due to risk premiums, compared to 3 to 5 per cent for proven technologies. Over a decade-long timeline, those percentage points compound significantly.
The path from FOAK to NOAK requires sustained deployment of identical designs. South Korea’s nuclear programme built 24 reactors to achieve meaningful cost reductions. For SMRs, industry projections suggest reaching the learning curve plateau after 5 to 7 units or 10 to 20 gigawatts of installed capacity. That’s a massive deployment that may take decades—if it happens at all.
If you’re evaluating SMR power purchase agreements, understanding FOAK versus NOAK assumptions matters. If a vendor quotes $60 per megawatt-hour LCOE, that’s almost certainly NOAK pricing assuming hundreds of future deployments. FOAK reality, as NuScale demonstrated, runs significantly higher.
When will SMRs actually be operational for data centre deployment?
The industry narrative says “2030,” which sounds close. But that timeline assumes everything goes perfectly. Nuclear projects historically do not go perfectly.
Everyone’s targeting the 2030 to 2035 window. Google’s agreement with Kairos Power targets 2030. TerraPower‘s Natrium reactor broke ground in June 2024 with a 2030 target. GE Hitachi’s BWRX-300 in Canada targets 2029.
NuScale achieved design certification from the NRC in August 2020—the first and so far only SMR design certified in the United States. The Idaho project was originally projected to be operational by 2024. Instead, it was cancelled in 2023 before construction even began.
For vendors still working towards design certification, the timeline looks like this:
- Design certification: 3 to 5 years of NRC review
- Construction permit: 2 to 3 years for site approval
- Construction: 3 to 5 years (claimed timeline; unproven for FOAK)
- Commissioning: 1 to 2 years of testing
That’s 9 to 15 years from today for vendors without design certification, assuming no delays. Vogtle Units 3 and 4 in Georgia cost $35 billion and were completed seven years behind schedule—and those were proven conventional reactor designs.
If you’re planning data centre expansions between 2025 and 2028, SMRs are not viable. Renewable PPAs are available immediately with construction timelines of 1 to 2 years. Even if everything goes perfectly, SMR power is 5 to 10 years away minimum.
What is the difference between FOAK and NOAK costs, and when will SMRs reach NOAK pricing?
FOAK—First-of-a-Kind—costs reflect what it takes to deploy initial commercial units. You’re paying for engineering refinement, supply chain development, regulatory learning, and all the unexpected challenges when moving from design to reality.
NOAK—Nth-of-a-Kind—costs represent what happens after you’ve built enough identical units to work out the kinks. Supply chains mature. Construction crews gain experience. Costs drop significantly.
The gap is where SMR economics live or die.
Industry projections assume FOAK costs are 2 to 3 times higher than eventual NOAK pricing. Based on NuScale’s experience, that gap may be larger—cost escalation suggests FOAK reality can be 3 to 4 times higher than initial estimates.
So when do SMRs reach NOAK pricing? Optimistically, the 2040s—assuming successful FOAK deployments in the 2030s and sustained build-out thereafter.
South Korea built 24 reactors to achieve meaningful cost reductions. For SMRs, projections suggest reaching the plateau after 5 to 7 units or 10 to 20 gigawatts of installed capacity. At current deployment rates, that’s 15 to 20 years away minimum.
The Department of Energy estimates lifetime cost could drop by around 70 per cent to $60 per megawatt if designs become standard and are repeatedly produced at scale. But getting there requires hundreds of identical units, which raises a challenge: vendor fragmentation.
Currently, you have NuScale, Kairos Power, TerraPower, X-energy, and GE Hitachi all pursuing different designs. Each vendor is starting their own FOAK journey. None can achieve economies of series production whilst the market is fragmented.
Compare this to renewables. Solar costs dropped 89 per cent between 2010 and 2020. Wind costs fell 70 per cent. That’s what learning curves look like at scale. SMRs need similar deployment volumes, but they’re starting from a fragmented base of competing designs.
For data centre planning: NOAK pricing is 15 to 20 years away minimum. Any decision today must be based on FOAK economics, not aspirational projections that may never materialise.
Why did the NuScale project in Idaho get cancelled and what does it mean for SMR economics?
The NuScale Idaho project was supposed to be the proof of concept that validated SMR economics. It had everything: the first and only NRC-certified SMR design in the United States, over $1.4 billion in DOE support committed, a consortium of public utilities ready to buy the power, and a proven site.
It didn’t work. In November 2023, the project was cancelled due to cost increases.
In 2020, the projected cost was $3.6 billion for 720 megawatts. By 2023, that had escalated to $9.3 billion for just 462 megawatts—less power at nearly three times the cost. That’s over $20,000 per kilowatt, compared to the original $5,000 per kilowatt estimate.
The original LCOE target was $58 per megawatt-hour. By January 2023, even with subsidies, the target price was $89 per megawatt-hour. Based on the final $9.3 billion cost, actual LCOE would have exceeded $100 per megawatt-hour even with subsidies.
Participating utilities could buy cheaper power elsewhere. Why lock into expensive 40-year nuclear contracts when alternatives cost half as much?
The implications are significant. NuScale did everything right: achieve design certification, secure government support, line up customers, select a proven site. And still, FOAK economics killed the project.
What does this mean for other SMR vendors? They all face the same FOAK challenges NuScale couldn’t overcome: engineering changes, supply chain development, construction inefficiencies, regulatory learning curves, higher financing costs, and competition from renewables.
If you’re evaluating SMR power purchase agreements, the NuScale cancellation is a cautionary tale. If the most advanced, best-funded SMR project couldn’t achieve viable economics, what makes the next project different? Scrutinise cost assumptions carefully. Understand whether pricing reflects FOAK or NOAK economics.
What are the main SMR vendors and how do their costs and timelines compare?
The SMR vendor landscape is crowded with competing designs, each facing the same FOAK economics that killed NuScale’s Idaho project.
NuScale achieved the first NRC design certification in August 2020. The Idaho project was cancelled in 2023 after costs ballooned to $9.3 billion.
Kairos Power signed a Master Plant Development Agreement with Google to develop 500 megawatts with the first unit targeted by 2030. Design certification is still in progress.
TerraPower, backed by Bill Gates, broke ground on its Natrium reactor in Wyoming in June 2024. The $4 billion project targets 345 megawatts by 2030.
X-energy develops a high-temperature gas-cooled reactor. Amazon signed an agreement for a four-unit, 320-megawatt project, alongside a Dominion Energy partnership for 300 megawatts. Amazon invested $500 million in X-energy, targeting 5 gigawatts by 2039.
GE Hitachi’s BWRX-300 received construction approval for the Darlington site in April 2025. The project costs CAD $7.7 billion with a 2029 target. This represents approximately $25,000 per kilowatt—well above industry projections and similar to NuScale’s escalated costs.
Here’s the challenge: each vendor has a different design using different technologies. Tech giants have committed over $10 billion to nuclear partnerships with 22 gigawatts of projects in development globally, but they’re not working towards a standardised design—something experts recommend.
This fragmentation means each vendor is separately working through FOAK challenges. None can share manufacturing supply chains, construction learning curves, or regulatory experience.
Most vendors release aspirational NOAK projections rather than realistic FOAK pricing. When you see a vendor claiming $60 per megawatt-hour LCOE, ask: is that FOAK or NOAK? How many units need to be deployed to reach that price? NuScale had NRC certification and still couldn’t make FOAK economics work.
The vendor landscape suggests a fragmented market with no clear winner, all facing similar FOAK challenges, and timelines realistically pushing into the 2030s at minimum. Compare this to proven reactor designs where engineering is settled, and you see the gap between promise and delivery.
FAQ
Are SMRs cheaper than traditional nuclear power plants?
No. SMRs cost more per megawatt because they sacrifice economies of plant scale. First-of-a-kind SMRs cost 2 to 3 times more per kilowatt than proven large reactor designs. NuScale’s Idaho project escalated to over $20,000 per kilowatt before cancellation.
When will the first commercial SMR be operational in the United States?
Realistically 2030 to 2035, assuming no delays. This timeline includes design certification (3 to 5 years), construction permits (2 to 3 years), construction (3 to 5 years), and commissioning (1 to 2 years). Historical nuclear projects routinely experience multi-year delays.
How does SMR electricity cost compare to renewable energy PPAs?
Renewable PPAs offer solar at $30 to $50 per megawatt-hour and wind at $25 to $45 per megawatt-hour, available immediately. SMR LCOE projections claim $60 to $80 per megawatt-hour but real-world projects indicate over $100 per megawatt-hour, and won’t be operational until the 2030s.
What is FOAK financing and why does it increase SMR costs?
FOAK (First-of-a-Kind) financing addresses unique risks of deploying the first commercial unit of new nuclear technology. It requires government support and accepts higher interest rates (8 to 12 per cent). FOAK projects cost 2 to 5 times more than eventual NOAK (Nth-of-a-Kind) pricing.
Why was the NuScale Idaho project cancelled?
Costs escalated from $3.6 billion to $9.3 billion between 2020 and 2023, making electricity prices (over $100 per megawatt-hour) uncompetitive with renewables. Despite over $1.4 billion in DOE support and NRC design certification, FOAK economics made the project unviable.
What government support is available for SMR projects?
The DOE allocated $452 million for SMR licensing support and offers loan programmes for first-of-a-kind nuclear projects. In December 2025, DOE selected TVA and Holtec for up to $800 million. However, NuScale’s cancellation despite $1.4 billion in DOE commitments shows government financing alone cannot overcome cost challenges.
Which companies are building SMRs and when will they be ready?
Major vendors include NuScale (Idaho project cancelled 2023), Kairos Power (Google PPA, targeting early 2030s), TerraPower (Bill Gates-backed, 2030 timeline), X-energy (Amazon backing, 2030s target), and GE Hitachi BWRX-300 (Ontario project targeting 2029). All face 2030 to 2035 timelines minimum.
Can SMRs power data centres more cost-effectively than renewables?
Not in the near term. SMRs won’t be operational until the 2030s with FOAK costs making LCOE over $100 per megawatt-hour. Renewable PPAs are available today at $30 to $50 per megawatt-hour. Even accounting for storage costs, renewable combinations often remain more cost-effective.
For a comprehensive overview of all aspects of the nuclear-AI intersection, including company strategies and regulatory frameworks, see our complete guide to Big Tech’s nuclear power pivot.
What is the difference between FOAK and NOAK costs for SMRs?
FOAK (First-of-a-Kind) costs include engineering refinement, supply chain development, and regulatory learning for initial deployments. NOAK (Nth-of-a-Kind) costs reflect reductions after hundreds of identical units achieve manufacturing scale. SMR projections often cite NOAK targets ($60 to $80 per megawatt-hour) whilst FOAK reality is 2 to 3 times higher.
How long does NRC approval take for SMR designs?
Design certification requires 3 to 5 years of NRC review. Following certification, each project needs a construction permit (2 to 3 years) and operating licence. Total regulatory timeline from design certification to operational approval spans 5 to 8 years minimum.
Do SMRs require less construction time than traditional nuclear plants?
SMRs claim 3 to 5 year construction timelines versus 7 to 10 years for large reactors. However, this advantage is unproven for FOAK deployments. Total project timeline still spans 8 to 15 years from today for vendors without design certification. For data centre planning, this offers no advantage over renewables deployable in 1 to 2 years.
