Seven sites, zero reactors under construction, and a policy that has now survived one federal election loss without updating its central numbers. That’s roughly where the Coalition’s nuclear plan sits heading into summer 2026 – still the same seven former coal sites first named in mid-2024, still built around CSIRO’s GenCost estimates as the thing everyone argues about rather than agrees on. Agree on the units first, as I keep telling people at dinner parties who really didn’t ask: are we talking overnight capital cost, levelised cost of energy, or total build cost including financing? The debate usually collapses because three different numbers get used interchangeably.
I’ve gone back through the Coalition’s own policy documents and the subsequent commentary from the CSIRO, the Clean Energy Council and a few engineering firms with actual large-reactor experience, and the honest position is that nothing has moved on the fundamentals since the plan was first released. The arguments are the same ones from eighteen months ago, just with another year added to however long the construction timeline is supposed to be.
What the plan actually says, as published #
The policy, in its public form, nominates seven sites for large-scale nuclear: Liddell and Mount Piper in NSW, Loy Yang in Victoria, Tarong and Callide in Queensland, and Collie and Muja in Western Australia. All are current or former coal sites, chosen partly for existing transmission connections and partly, you’d assume, for the political convenience of pointing at a site that’s already losing a coal plant rather than opening a brand new greenfield fight.
The stated logic is twofold. First, that nuclear provides firm, dispatchable baseload that renewables plus storage can’t match on a like-for-like basis. Second, that using existing coal sites saves on transmission costs because the connection infrastructure is mostly already there. Both claims have some truth in them. Neither resolves the two questions that actually matter for anyone trying to plan around this: when would power from these plants actually reach the grid, and what would it cost per megawatt-hour by the time it does.
The timeline problem nobody in the debate wants to own #
Here’s where I think both sides of this argument get a bit sloppy. The pro-nuclear case tends to point to South Korea’s APR1400 build times – around seven to eight years from first concrete to grid connection in a mature domestic programme, and implicitly assumes Australia could match that. The anti-nuclear case tends to point to Hinkley Point C in the UK, which has blown out from an original 2017 completion target to sometime past 2029, and treats that as the only comparable precedent.
Both comparisons miss the actual starting point. Australia has no domestic nuclear regulator with reactor licensing experience, no established supply chain for reactor-grade components, no trained construction workforce for this kind of build, and currently a federal moratorium under the Australian Radiation Protection and Nuclear Safety Act 1998 and corresponding state-level prohibitions that would need to be unwound before a shovel goes in the ground. The CSIRO’s GenCost report, now in its latest annual update, puts first-of-a-kind large reactor commissioning in Australia at the earliest in the mid-2040s, accounting for the regulatory build-out alone. That’s not a renewables-lobby number. It’s the standard government-commissioned cost benchmarking exercise that both the AEMO Integrated System Plan and most utility investment decisions already lean on.
I’ll admit I used to think the mid-2040s estimate was conservative to the point of being a political statement itself. Having now sat through a few briefings on what licensing regimes in South Korea, the UAE and France actually took to establish from scratch, I’ve come around to thinking it’s roughly right, maybe even generous if Australia tries to build genuine domestic regulatory capacity rather than importing it wholesale.
Small modular reactors don’t solve the timeline, they change its shape #
The plan leaves room for small modular reactors as a faster-to-deploy alternative, and this is where I’d gently push back on both camps again. SMRs are frequently spruiked as the answer to the large-reactor timeline problem, but as I’ve written before, the actual state of SMR technology globally is one of delayed demonstration projects and cancelled first-of-a-kind orders rather than a mature, buyable product. NuScale’s cancelled Utah project is the standing example of how the economics can fall over even with a willing customer and US federal backing. There’s no commercially operating SMR fleet anywhere in the world yet that Australia could simply order from a catalogue.
That doesn’t mean SMRs are a dead idea. It means treating them as a near-term timeline fix is wishful. If Australia wants SMRs in the 2030s it needs to be ordering from a vendor with an operating reference plant, and right now that list is short to the point of being empty for the designs under serious discussion here.
The cost numbers everyone quotes and why they don’t match #
GenCost’s 2024-25 update put large-scale nuclear’s levelised cost well above firmed renewables – solar and wind backed by storage and transmission – even before accounting for the first-of-a-kind premium Australia would pay as a country with no existing fleet. The Coalition and some supportive engineering voices have disputed the GenCost methodology, particularly its treatment of plant lifespan and capacity factor assumptions, and that’s a legitimate technical argument to have. CSIRO has published its methodology and responded to submissions each cycle, which is more transparency than either side protagonist tends to credit it with.
What’s missing from the Coalition’s own costing is a bottom-up capital cost figure with the same level of detail GenCost applies to renewables and storage. The policy has pointed to international reactor costs as a rough guide rather than publishing an Australian-specific build estimate with site remediation, grid connection upgrade, workforce training and financing costs itemised. Until that exists, comparing the plan’s implied cost to GenCost’s levelised cost is comparing an estimate to an assumption, not two estimates.
I don’t think that makes the nuclear case dishonest. Early-stage energy policy rarely comes with bankable costings – ask anyone who’s followed Borumba’s pumped hydro budget move around over the past two years, or Snowy 2.0’s long run of schedule revisions. But the absence of a comparable costing is a real gap, and treating “nuclear is cheaper than GenCost says” as settled without publishing the workings is where I think the pro-nuclear argument weakens itself.
What the sites would actually displace, and when #
The seven nominated sites matter because of what’s happening to them anyway. Liddell has already closed. Loy Yang and the Queensland sites are on retirement trajectories that, per AEMO’s published closure expectations, run through the early-to-mid 2030s. Muja in Western Australia is well into its own wind-down, a process I’ve covered in some detail alongside the Collie battery projects now filling part of that gap.
If large nuclear reactors can’t realistically reach commissioning before the mid-2040s, there’s a fifteen-to-twenty-year gap between when these coal sites actually leave the grid and when nuclear could plausibly replace them at any of the nominated locations. That gap has to be filled by something, and right now the something being built in its place (batteries, gas peaking plant, transmission and renewables) is exactly what the Integrated System Plan already assumes gets built regardless of what happens with nuclear policy. Which raises the obvious question the Coalition’s plan has never fully answered: what runs these sites in the gap years, and does that interim buildout get stranded or does it simply become the permanent answer by default?
My own read, for what it’s worth, is that the gap-filling buildout becomes permanent by sheer weight of sunk capital and operating contracts, nuclear or no nuclear. Once a battery fleet, a REZ transmission spine and a decade of operating experience exist at a site, ripping that out for a reactor that arrives in the 2040s becomes a much harder sell to whichever government is actually holding the budget at the time, regardless of which party wrote the original policy.
The fair comparison nobody’s running yet #
The debate that would actually be useful is nuclear versus firmed renewables on a proper like-for-like basis: same delivery date, same capacity factor assumptions, same financing cost of capital, published side by side. That comparison has been attempted in pieces – GenCost does some of it, the AEMC and various engineering consultancies have done fragments, but nobody has published the full apples-to-apples version that would let a reasonably informed reader actually judge the claim that nuclear is cheaper once you strip out ideology on both sides.
Until someone does that properly, the debate will keep running the way it has for the better part of two years: proponents quoting Korean build times against a renewables-plus-storage cost base, opponents quoting Hinkley Point delays against a nuclear cost base, and precious little meeting in the middle. It’s not a maths problem anymore. It’s a data-transparency problem, and both sides have reasons not to fix it.
Where that leaves the policy debate #
None of this means nuclear is impossible for Australia, and I’d resist anyone who tells you the physics or engineering rules it out. Countries smaller and less industrially developed than Australia have built domestic nuclear fleets from a standing start. It means the current policy, as published, is a site list and a set of assumptions rather than a costed, dated project. That’s a reasonable starting point for a policy document eighteen months before an election. It’s not yet a basis for comparing against the firmed-renewables pathway the grid is actually building right now, which is why so much of this argument keeps circling the same four or five numbers without landing anywhere new.
The grid doesn’t wait for policy certainty, unfortunately. Coal units keep closing on their own schedules regardless of what’s decided about what replaces them two decades out, and the generation gap in between gets filled by whoever actually shows up with capital and a connection agreement. On current form, that’s not reactors.
– Tom Fitzgerald, Baseload & Fuels Correspondent
Photo by Jason Mavrommatis on Unsplash