SMR vs. large nuclear reactor: which is actually cheaper?
The marketing says small reactors save money, but the actual numbers tell a messier, more interesting story
Every SMR pitch deck has the same slide. Factory-built modules, shorter construction timelines, lower financing risk, cheaper power. It sounds like common sense: smaller things generally cost less. 💰 But nuclear power has never played by common sense economics, and the honest answer to “which is cheaper” is genuinely more complicated than either side of this debate wants to admit. Let’s actually run the numbers. 🔬
The economies-of-scale problem nobody can escape
Here’s the physics-meets-economics reality that every SMR company has to fight against: bigger reactors have always been cheaper per unit of power, because a lot of the cost, containment structures, control systems, safety equipment, doesn’t shrink proportionally with reactor size. This is why the industry spent seventy years building reactors bigger, not smaller. 📉
The current cost estimates make that gap pretty explicit. <cite index=”65-1”>The EIA estimates overnight capital costs for SMRs at around $9,500 per kilowatt, actually higher than large reactors on a per-kilowatt basis</cite>, while <cite index=”65-1”>a large light-water reactor comes in at about $7,800 per kilowatt, or roughly $8,100 once regional cost differences get factored in</cite>. That is not a small gap, and it runs directly against the “smaller is cheaper” pitch.
South Korea’s SMART SMR: <cite index=”66-1”>projected at $10,000 per kWe, about four times higher than the APR1400, a large reactor from the same country</cite>
EU comparison: <cite index=”67-1”>the IEA estimates SMR overnight costs around $10,000 per kW in Europe, versus $6,600 per kW for traditional nuclear</cite>
Academic bottom-up modeling: a detailed light-water SMR study found <cite index=”63-1”>an overnight cost of $4,844 per kW, with a levelized cost of $89.6 per megawatt-hour</cite>, closer to parity but still not obviously cheaper
The pattern across nearly every independent estimate is the same. On a straight per-kilowatt basis, first-of-a-kind SMRs are not cheaper than large reactors. Some estimates put them meaningfully more expensive. 🚨
NuScale’s cautionary tale
If you want a real-world case study instead of a spreadsheet projection, look at what happened to the industry’s most advanced SMR project. It is not pretty, and it is the example every skeptic reaches for first. ⚡
<cite index=”62-1”>NuScale and the Utah Associated Municipal Power Systems announced that costs for their 462-megawatt SMR project had risen dramatically, with the target power price climbing from $58 to $89 per megawatt-hour, a 53% increase</cite>. The construction estimate jumped even harder. <cite index=”62-1”>Estimated construction costs rose 75%, from $5.3 billion to $9.3 billion</cite>, which <cite index=”62-1”>put the project at roughly $20,139 per kilowatt, about as expensive as the Vogtle mega-reactor project in Georgia</cite>. The project was eventually cancelled outright. Vogtle itself is no poster child either. <cite index=”65-1”>Its two new reactors came online in 2023 and 2024 at a total cost exceeding $30 billion, against an original budget of roughly $14 billion, working out to about $13,600 per kilowatt</cite>.
So the uncomfortable truth is this: both SMRs and giant reactors have a track record of blowing their budgets. Small reactors were supposed to be immune to this because factory production avoids the on-site chaos that sank Vogtle. NuScale’s collapse suggests that promise hasn’t been proven yet, at least not at first-of-a-kind scale. 🌱
Where SMRs might actually win
Okay, so the pessimistic case is loud and well documented. But it’s not the whole picture, and dismissing SMR economics entirely would be its own kind of oversimplification. There are a few places where the math genuinely favors small reactors. 📈
Financing risk: <cite index=”61-1”>capital costs on a dollar-per-kilowatt basis may be higher for smaller units due to lost economies of scale, but there may be real advantages in financing costs from shorter project durations and lower interest exposure</cite>
Construction speed: one bottom-up analysis found <cite index=”63-1”>a light-water SMR construction duration averaging 4.5 years with a 90% probability of landing between 3.4 and 6.0 years</cite>, versus five-plus years typical for large plants
Learning curve potential: <cite index=”59-1”>Idaho National Laboratory projects a high-case scenario where deploying 32 SMR units drives a 55.6% reduction in overnight construction cost through manufacturing learning effects</cite>
Reduced overrun risk: <cite index=”68-1”>a single-unit SMR requires significantly less total on-site labor, and if built by an experienced workforce, it could avoid the cost-overrun risks tied to megaprojects</cite>
That learning curve is really the whole ballgame. Nobody disputes that a tenth reactor off an assembly line should cost less than the first one. The argument is entirely about how fast that curve bends, and whether investors are willing to eat the expensive early losses to get there. <cite index=”60-1”>Wood Mackenzie expects first-of-a-kind SMR costs around $180 per megawatt-hour, dropping 40% to about $100 by 2030 as manufacturing scales up</cite>. That’s a big bet on a curve that, right now, exists mostly on paper.
The FOAK versus NOAK trap
This distinction, first-of-a-kind versus nth-of-a-kind, is where most of the confusion in this whole debate actually lives, and it’s worth being blunt about it. Nearly every rosy SMR cost projection you’ll see is an NOAK number, not a FOAK number. Nearly every real-world cost overrun story, NuScale included, is a FOAK story. Comparing the two is comparing a company’s five-year sales forecast to its first quarter of actual revenue. 🧬
FOAK reality: high costs, unproven supply chains, first-time regulatory approval, workforce retraining, all the expenses of doing something for the first time
NOAK promise: standardized designs, established suppliers, trained crews, manufacturing efficiencies compounding across dozens of units
The gap between them: <cite index=”67-1”>even under the IEA’s optimistic Announced Pledges Scenario, cost parity between SMRs and conventional nuclear might not arrive until mid-century, and under the more conservative Stated Policies Scenario, SMR capital costs stay roughly one-third higher even by 2050</cite>
<cite index=”67-1”>As of 2024, only three SMRs had actually been constructed globally, each in a different country, which is nowhere near enough real-world experience to prove the learning curve theory one way or the other</cite>. Everything past that is projection, and projections in this industry have a rough track record. If you’re trying to separate genuine NOAK progress from optimistic marketing across the dozens of SMR companies now raising capital, that’s precisely the kind of deal-by-deal tracking SMRbrief Pro exists to sort through.
So which one actually wins?
Right now, on hard numbers, large reactors hold a real edge per kilowatt, and anyone telling you SMRs are simply cheaper today is skipping past a lot of inconvenient data. But “today” is doing a lot of work in that sentence. The entire SMR bet is that serial manufacturing, shorter build times, and lower financing risk eventually flip the equation, the same way factory production flipped costs in solar panels and batteries. Whether nuclear behaves like those industries, or whether reactors are just too complex and too regulated to follow the same curve, is genuinely unresolved. 🚀
I lean skeptical of near-term parity, mostly because the industry has promised this before and Vogtle happened anyway. But I also think dismissing the whole SMR category because of one cancelled Utah project is too easy. What would actually change your mind here, a handful of on-time, on-budget NOAK deployments, or is the learning curve theory itself the thing that needs more scrutiny before another dollar of public money goes toward proving it?