NuScale’s first US customer, a group of Utah municipal utilities called UAMPS, walked away from its small modular reactor project in late 2023 after the targeted cost per megawatt-hour had roughly doubled from the original estimate. That project had government backing, a willing customer base, and a design that had already cleared the US Nuclear Regulatory Commission. It still didn’t survive contact with an actual budget. I bring this up not to pile on NuScale, which is still the only SMR vendor with a certified design in the United States, but because it’s the cleanest case study we have of what happens when small modular reactor economics meet a real procurement process rather than a slide deck.
Small modular reactors come up constantly in Australia’s energy debate, usually attached to the Coalition’s nuclear policy or to mining companies eyeing remote power supply. The pitch is seductive: factory-built reactors, trucked to site, assembled faster and cheaper than the giant plants of the 1970s and 80s, scaled to whatever a grid or a mine actually needs. Agree on the units first, though, because “small modular reactor” covers a spread of designs from 300 megawatts down to a few megawatts, at wildly different stages of maturity, and conflating them is how these conversations go sideways.
What “modular” is actually supposed to buy you #
The theory behind SMRs isn’t exotic. Large reactors are expensive partly because every one is close to a bespoke civil engineering project, built on site, with first-of-a-kind cost overruns baked into nearly every Western nuclear build this century. Plant Vogtle in Georgia, the most recent US example, ran roughly double its original budget and years over schedule. The SMR answer is to shrink the reactor enough that core components can be manufactured in a factory under controlled conditions, shipped by road or rail, and assembled on site with far less bespoke concrete pouring and far more repeatable engineering.
That’s a genuinely sound idea if you can build enough of them. The economics depend on learning curves: unit ten should be cheaper than unit one because the factory and the workforce get better at it, the way battery and solar manufacturing have driven down costs through scale. The catch is you need a committed order book of dozens of units to get anywhere near that curve, and right now there isn’t a vendor on earth with dozens of commercial units operating, under construction, or even firmly contracted. We’re still at the point where every SMR built is close enough to first-of-a-kind that the promised modularity hasn’t had a chance to pay off.
Where the designs actually stand #
NuScale remains the reference case because it’s furthest through regulatory approval, with its 50-megawatt module design certified by the NRC. But the commercial pipeline is thin: the UAMPS cancellation left it without a flagship US customer, and its more recent activity has shifted toward international interest and smaller demonstration-scale proposals rather than firm multi-unit orders.
X-energy, backed partly by Amazon, is pursuing a high-temperature gas-cooled design aimed at industrial heat customers as much as grid power, with a planned demonstration at a Dow chemical site in Texas. TerraPower, Bill Gates’s venture, is building its Natrium sodium-cooled design in Wyoming on the site of a retiring coal plant, explicitly trading on coal workforce transition: a model Australia’s own coal communities would recognise immediately, since it’s the same logic behind proposals to put reactors on retiring coal sites here. Rolls-Royce SMR in the UK is probably the most advanced non-US large-scale SMR programme, targeting a roughly 470-megawatt design that’s really a small conventional reactor rather than the micro-reactor end of the category, and it has UK government backing through the Great British Nuclear process. None of these has a unit generating commercial power into a grid today.
China and Russia are ahead on actual operation – China’s Linglong One demonstration unit and Russia’s floating Akademik Lomonosov plant are running, but both are state-directed programmes with financing, siting, and regulatory conditions that don’t translate cleanly to how Australia would need to build.
The regulatory starting line Australia hasn’t crossed #
Australia currently has a federal legislative prohibition on nuclear power generation, dating from amendments made in the late 1990s to the Australian Radiation Protection and Nuclear Safety Act and the Environment Protection and Biodiversity Conservation Act. The Coalition’s nuclear policy, covered in detail in our piece on whether Australia actually needs nuclear power, would need that ban lifted federally and in several states, plus the construction of an entire regulatory apparatus: a nuclear safety regulator with the depth of the NRC or the UK’s Office for Nuclear Regulation doesn’t exist here and can’t be stood up in a term of government. The CSIRO and AEMO’s own planning work, including the Integrated System Plan, has consistently treated nuclear as outside the feasible build envelope for the 2030s on both cost and timeline grounds, a view we’ve tracked in our look at nuclear versus firmed renewables and the cost maths.
That’s a genuinely separate question from whether the technology works. A design can be sound engineering and still be fifteen years away from a connection agreement in the NEM, once you stack regulatory stand-up, site approval, first-of-a-kind construction risk, and the fact that nobody in Australia has built a reactor of any size since Lucas Heights, which is a research reactor, not a power station.
The cost curve that keeps moving the wrong way #
Every serious cost estimate for first-generation SMRs has moved up, not down, over the past five years, which is the opposite of what the modularity thesis predicted at this stage. The CSIRO’s GenCost report, produced annually with AEMO, has consistently put nuclear, large or small, well above firmed renewables on levelised cost, and widened rather than narrowed that gap in recent editions as actual Western nuclear construction costs have kept disappointing optimists. GenCost isn’t an anti-nuclear document; it’s a reasonably dry exercise in taking announced and actual project costs and running them through a standard methodology, which is precisely why the trend line matters more than any single year’s number.
None of this means SMRs are a dead idea everywhere. The case for them is strongest in applications where a 300-megawatt coal or gas plant equivalent, or something a tenth that size, is genuinely the right unit of power: isolated industrial sites, remote mining operations, or small grids like parts of WA’s SWIS network that can’t easily import power from elsewhere. Those are exactly the settings where the economics of building one giant reactor never worked anyway, and where a factory-built, truck-delivered unit has a real argument if the price ever lands where the vendors promise. It just hasn’t landed there yet, for anyone, in the West.
Why this keeps getting compared to batteries and gas peakers, badly #
A reactor is not a peaking asset. SMRs, like large nuclear plants, run most efficiently as baseload, cycling output slowly if at all, which is a different job to the one most of the NEM’s firming capacity gap is actually trying to fill. Our explainer on what firming means and why gas keeps coming up covers this in more detail, but the short version is that the grid’s current stress point is the evening ramp and short-duration gaps when solar drops off and wind is soft, not a shortage of always-on generation per se. A fleet of SMRs sited to replace retiring coal could, in principle, do a similar structural job to coal, but so could gas peakers and big batteries, both of which exist today, are approved under known regulatory frameworks, and connect to the grid on timelines measured in months to a few years rather than a decade-plus.
I’ll be fair to the nuclear case on one point that often gets lost: land use and transmission. A reactor replacing a coal unit at Liddell or Bayswater would reuse existing transmission connections and a lot of existing workforce skill, which is a real advantage over building fresh renewable energy zones and the transmission to match, a problem we’ve covered at length in NSW’s coal exit and REZ rollout gap. That’s a legitimate point in the SMR column. It just doesn’t solve the cost and timeline problem on its own.
My read on where this actually sits #
I reckon the contrarian view isn’t “SMRs will never work” – plenty of credible engineers think the core physics and manufacturing logic are sound given enough committed orders. The contrarian view, if there is one, is that the technology risk is smaller than the commercial and regulatory risk, and almost nobody in the Australian debate talks about it that way. The design side is progressing steadily, if more slowly than vendors promised five years ago. What’s missing isn’t an engineering breakthrough; it’s a customer willing to sign for the fifth or tenth unit before the first one has proven itself, because that’s the only way the learning curve the whole pitch depends on actually shows up. UAMPS was that customer in Utah, and it walked. Finding an equivalent here, inside a market that doesn’t yet legally permit nuclear generation, is a bigger ask than the brochures suggest.
If I had to bet cider money on it, and Perth’s Measure for Measure usually has something worth betting over – I’d say Australia sees serious SMR proposals attached to specific mine sites or industrial precincts well before it sees one connected to the NEM. That’s a narrower, less glamorous outcome than the political debate implies, but it’s probably the honest shape of where this technology is actually heading over the next decade.
– Tom Fitzgerald, Baseload & Fuels Correspondent
Photo by Sherise Van Dyk on Unsplash