Why SMR Projects Keep Running Over Budget — And What's Changing
The pattern of cost blowouts in small modular reactor projects isn't random bad luck — it's structural, and the industry finally knows it.
Nuclear energy has a cost problem that no amount of enthusiasm from Goldman Sachs, Microsoft, or the UK government seems to solve. Pick almost any SMR project from the last decade, follow it for a few years, and watch the budget estimates balloon. It’s almost hypnotic, in a horrifying sort of way. The NuScale flagship project in Idaho went from $5.3 billion to $9.3 billion before collapsing entirely. Argentina’s CAREM reactor has racked up cost overruns of around 700% since construction began in 2013. Russia’s floating reactor, the Akademik Lomonosov, blew its budget by roughly 400%. China’s pebble-bed HTR-PM came in at about 300% over.
These aren’t outliers. They’re the entire dataset.
For an industry promising a cheaper, faster, more manageable alternative to the nuclear megaprojects of the 20th century, this is a genuinely uncomfortable place to be. So why does it keep happening? And — more interestingly — what, if anything, is actually starting to change?
The structural reasons cost overruns happen
Start with something called the iron law of megaprojects. According to research cited in a techno-economic analysis published in Applied Energy, nine out of ten large infrastructure projects go over budget, with cost overruns exceeding 50% not being uncommon. SMRs were supposed to sidestep this trap through their smaller, more manageable scale. That theory makes intuitive sense. What it missed is that being small doesn’t automatically make you simple. 🏗️
The structural cost drivers are stubborn:
Regulatory complexity: Each reactor design needs its own licensing process, and in the US that process can eat years and hundreds of millions of dollars before a shovel enters the ground. NuScale spent roughly $1.8 billion just getting to the point of design certification — with unresolved safety questions still on the table.
Supply chain immaturity: You can’t order SMR components the way Boeing orders aircraft parts. The supply chain is thin, bespoke, and expensive. Commodity price swings — steel, concrete, specialized equipment — hit nuclear projects harder than almost any other construction type.
Workforce gaps: Skilled nuclear construction workers are rare, expensive, and don’t scale quickly. Every first-of-a-kind project trains its workforce essentially from scratch. 👷
Design churn: Many projects have changed their reactor designs mid-development, which resets timelines and explodes budgets. NuScale’s original 50-megawatt design was upsized to 77 megawatts when the economics of the smaller unit failed to hold up.
Financing costs: Nuclear projects tie up capital for a decade or more before generating a single watt of electricity. Every year of delay is a year of interest payments on top of construction costs.
According to IEEFA’s May 2024 report, the pattern isn’t improving. The data from the few SMRs that have actually been built shows cost outcomes that look uncomfortably like large-reactor failures, just at a smaller scale.
The FOAK trap — and why it’s real
There’s a concept the nuclear industry uses to explain away first-project failures: FOAK, or First-Of-A-Kind. The argument goes that first plants are always expensive; the savings kick in once you’re building the 5th, 10th, or 20th unit — what engineers call NOAK (Nth-Of-A-Kind). This argument isn’t wrong exactly. But it’s being used to wave away a genuine problem. 🔬
According to the International Energy Agency’s 2025 projections, SMR overnight costs in the EU run around $10,000 per kilowatt, versus $6,600 per kilowatt for conventional nuclear. Research on FOAK-to-NOAK transitions suggests cost reductions of 20–30% are realistic with manufacturing learning curves and supply chain optimization — but only if:
Designs are standardized across many builds
Production volumes reach at least 30 to 50 units of the same reactor type
Supply chains are built and coordinated in advance
Regulatory processes are streamlined and consistent
Lars Thurmann-Moe of Arthur D. Little, speaking to NucNet in July 2025, put it plainly: “We need a single or small group of standardised designs to reach production volumes of at least 30 to 50 units to fully benefit from economies of scale.” That’s the catch. Right now, the IAEA counts over 80 different SMR designs in various stages of development globally. Eighty designs. For a market that hasn’t yet proven demand for one. The FOAK learning curve only works if you actually build enough reactors of the same design to move past FOAK. With this many competing designs splitting a limited pool of customers, that’s genuinely hard to achieve.
Do you think the proliferation of competing SMR designs helps or hurts the industry’s case for government support? It’s worth thinking through.
The financing trap — where the budget pressure actually lives
Here’s the part that doesn’t get enough attention: the overnight construction cost is only part of the story. Nuclear projects are unique in how brutally they’re punished by financing costs. A project that takes 15 years to complete pays interest on borrowed capital for 15 years. A project that takes 5 years pays for 5. The math is unforgiving. 💰
This is where SMRs have a genuinely real advantage — in theory. Consider the comparison laid out in a June 2025 GLOBSEC analysis: even with a higher overnight cost of $10,000/kW versus $6,600/kW for a conventional plant, an SMR completing in 5 years has a lower all-in cost than a conventional reactor taking 15 years, assuming a 5% financing rate. The numbers work out to roughly $12,763/kW for the SMR versus $13,721/kW for the conventional plant.
But — and this is a very significant but — the same analysis shows that extending the SMR construction timeline by just two years, from 5 to 7 years, is enough to wipe out that advantage entirely. This is why schedule overruns are so damaging. It’s not just the extra construction cost. It’s the compounding interest on every month of delay.
The financing trap has a few distinct components:
Private capital is scarce: Unlike bridges or pipelines, SMRs don’t generate revenue until fully operational, creating a decade-long cash-negative phase that most private investors won’t accept.
Cost-of-capital premium: Nuclear projects attract higher discount rates because of their risk profile, making the effective cost of every dollar borrowed even higher.
Government subsidy dependence: NuScale’s $89/MWh target price would have been significantly higher without $4 billion in federal subsidies, including a $1.4 billion Department of Energy contribution and a $30/MWh Inflation Reduction Act benefit. Strip those out and the economics collapse.
What’s actually changing — and why it’s not nothing
Enough bad news. Something is shifting, even if it’s slower than the press releases suggest. ⚡
The most credible change is the movement toward genuine factory-based manufacturing. GE Hitachi’s BWRX-300, which received construction approval at Ontario Power Generation’s Darlington site in April 2025, is designed with roughly 90% of its components manufactured in factory conditions, with on-site work limited to assembly. BWXT has already been contracted to manufacture the reactor pressure vessel — an early signal that the supply chain is starting to get real. GE Hitachi’s stated price target, if fleet production is achieved, is around $5,000 per kilowatt for the BWRX-300. That would represent a meaningful shift.
Rolls-Royce SMR, selected by the UK government through Great British Energy in June 2025 for domestic deployment, is targeting an LCOE below £70/MWh. The 470-megawatt pressurized water reactor design has a strong factory-manufacturing case, with the UK’s existing engineering base potentially giving it a genuine industrial advantage.
Regulatory reform is also moving, slowly:
The US Nuclear Regulatory Commission has been under pressure to streamline its licensing processes for advanced reactor designs
Canada’s CNSC gave formal safety approval for the Darlington BWRX-300 site in April 2025
The IAEA and OECD-NEA have been pushing for international regulatory harmonization, which would let a reactor certified in one country leverage that work in another
The Darlington project in Ontario is perhaps the most important near-term data point for the whole industry. Its first reactor, projected to cost CAN$6.1 billion (roughly US$4.4 billion), is scheduled for completion in 2030. If it comes in on time and on budget, it will be the first piece of real evidence that factory manufacturing and regulatory improvements can actually deliver what the models predict. If it doesn’t, the skeptics — and there are plenty of them — will have fresh ammunition. 🌱
The honest reckoning
The SMR industry is not facing a public relations problem. It’s facing an engineering and economics problem that no amount of enthusiasm from tech billionaires or government ministers makes easier. The cost overrun pattern is real, it’s documented, and it reflects structural challenges that take years — not quarters — to fix.
But writing off the entire technology because the first few attempts went badly is probably too hasty. Every energy technology goes through a painful first-mover phase. Solar and wind were expensive and unreliable for decades before learning curves and manufacturing scale made them competitive. The difference is that solar and wind didn’t require a decade-long regulatory process for each new installation.
The question the SMR industry has to answer honestly — not in investor decks, but in actual delivered projects — is whether factory manufacturing and standardization can do what advocates have promised for years. Darlington will tell us something. So will Rolls-Royce’s first UK unit, if it gets built. So will whatever comes next in the US, where TVA is expected to apply for a BWRX-300 license.
Watch the actual construction timelines. Watch whether the Darlington FOAK comes in at CAN$6.1 billion or blows past it. That number will tell you more about the future of SMRs than any projection ever will.