← Energy Realism

Why GenCost Gets Nuclear Wrong

CSIRO's GenCost 2024-25 is the single report every politician and journalist reaches for to say nuclear is too expensive for Australia — a headline A$173–288/MWh for nuclear against A$76–131/MWhfor firmed renewables. Here's the thing: GenCost doesn't fabricate that number. It buries the assumptions that decide it. Move them yourself below.

Recalculate GenCost yourself

Presets:
7%
GenCost: 7%
30 yr
GenCost: 30 yr
71%
GenCost: 53–89%
A$13,200/kW
GenCost: ~A$13,200 (post-IDC)
Nuclear LCOE — mature fleet (NOAK)
A$215/MWh
Above the firmed-renewable band.
…as a FIRST Australian build (+120% FOAK)
A$420/MWh
GenCost's own suggested first-of-a-kind premium — shown always, because hiding it is the dishonest move.
Firmed renewables (the fair like-for-like): A$76–131/MWh (wind + solar + storage + transmission, GenCost 2030) — notbare solar at A$35–63. The defaults reproduce GenCost's published A$173–288/MWh across the 53–89% capacity-factor band.

GenCost doesn't fabricate nuclear's cost — it's radically transparent and internally consistent. What it does is bury the assumptions that decide the answer: a 30-year capital-recovery period (its own analysis shows 60 years cuts the figure only ~10%, yet it headlined 30), a capacity-factor floor of 53% benchmarked to Australian coal dispatch rather than nuclear's ~90% capability, and a 7% discount rate that — on the most capital-heavy technology — does most of the work. Stack the defensible alternatives and a mature fleet lands near A$125/MWh, competitive with the top of firmed renewables. But honesty cuts both ways.

The assumptions that decide it

Four inputs do almost all the work. None of them is a faked number — each is a choice, and CSIRO chose the conservative end of every one. (Honestly: no single change makes nuclear cheapest — they only bite when stacked, and the biggest lever runs against nuclear. See the verdict below.)

Asset life (capital-recovery period)

GenCost uses
30 years in the headline
Defensible
60 years (US reactors are licensed to 60–80yr)

The marquee dispute — and the strongest defensible critique, because CSIRO's OWN sensitivity (Table 2-1) shows 60 years cuts large-scale LCOE only ~9–10%, yet it headlined 30. Why so small: a 60-yr loan racks up ~131% more interest at 7%, ~40% of LCOE is non-capital, and discounting makes years 31–60 nearly worthless. It only delivers a big cut PAIRED with a low discount rate.

GenCost 2024-25 Table 2-1; World Nuclear Association submission

Capacity factor

GenCost uses
53–89% (benchmarked to Australian COAL dispatch)
Defensible
90% (US fleet ~93%, global avg ~80%)

LCOE scales as 1/CF on the ~60% of cost that's capital — so the 53% FLOOR drives most of GenCost's wide spread. CSIRO concedes 89%→93% is only ~A$5/MWh, so the fight is the 53% floor: of the 51 reactors below 60% in 2023, most were French units returning from repairs (clean-2019 data: only ~5% genuinely <60%). The honest tension advocates must own: you can't claim BOTH 90% baseload AND nuclear-as-flexible-firming.

GenCost 2024-25 §2.2 & §6.2.5; WNA submission

Discount rate / WACC

GenCost uses
7% real, uniform across all technologies
Defensible
5% real (government / RAB-backed) — NOT 2–4%, which overstates it

The largest lever favourable to nuclear (>60% of its LCOE is financing) — but the WEAKEST to call 'bias'. A single uniform rate is the mainstream convention (NREL, AEMO) and arguably the FAIR, technology-neutral choice; it only disadvantages nuclear because nuclear genuinely IS the most capital-heavy option. Government financing lowers cost by SHIFTING overrun risk onto taxpayers, not removing it — and the same low rate would also cut renewables.

GenCost 2024-25 discount-rate section; IEA/NEA 2020

First-of-a-kind premium

GenCost uses
0% in the headline (NOAK); +120% suggested separately
Defensible
ON (+120%) for a first build; falls to 0% by the third

The most powerful lever — and it runs AGAINST nuclear. +120% roughly DOUBLES the capital component, dwarfing the ~10% from 60-yr life. Vogtle and Snowy 2.0 both ran ~100% over budget. An honest recalculator must never hide this. The one clean critic WIN here is SMR labelling: GenCost costs SMR at FOAK (~A$29,600/kW) while EVERY other tech is NOAK — vs US EIA's NOAK ~A$14,000/kW — an apples-to-oranges comparison CSIRO concedes it can't cleanly strip.

GenCost 2024-25 Table 3-1 (Flyvbjerg & Gardner 2023); IEEFA Sep 2024; US EIA 2024

What GenCost gets right

A critique you can't trust is worthless — so here's the other side, conceded plainly. GenCost is radically transparent (it publishes every workbook) and several of its choices are defensible. This is where the honest line falls.

“30-year life is rigged — reactors run 60–80 years.”
CSIRO: CSIRO ran the 60-yr case: it cuts large-scale LCOE only ~9–10% (100-yr ~10–12%), because 7% discounting heavily down-weights years 31–60 and ~40% of LCOE is fuel/O&M. 30 yr is the standard private capital-recovery period.
Verdict: Partly concede. The arithmetic is right — 60 years alone is only ~10%, so this single change does NOT make nuclear cheap. The legitimate win is narrow: CSIRO got ~10% and still headlined 30.
“The 53% capacity-factor floor is benchmarked to coal, not nuclear.”
CSIRO: Australia has zero nuclear history but 100+ years of coal; CF is a market-dispatch outcome, and nuclear in a ~2045 solar/wind-dominated NEM could be forced to ramp. The 89%→93% high end is only ~A$5/MWh.
Verdict: Partly concede on the LOW end; CSIRO wins the high end. The 53% floor is the real dispute; the 89% high is reasonable.
“A uniform 7% WACC penalises capital-heavy nuclear.”
CSIRO: A uniform real rate is 'the most informative and least distortionary approach' and the mainstream convention; a sub-market government rate shifts risk to taxpayers, and would also cut renewables' cost.
Verdict: Defend CSIRO. This is the weakest 'big' critique — the uniform rate is fair and technology-neutral. Don't call 7% 'rigged'; a hostile economist dismantles that.
“GenCost ignores the firming/integration cost of renewables.”
CSIRO: False as stated — CSIRO added itemised integration costs (storage, transmission, system security, spillage) and a full System LCOE from 2023-24; nuclear is still uncompetitive on it.
Verdict: Two-sided. Say 'may UNDER-count', not 'ignores'. The real gap: backup is sized on only 9 weather years (2011–19); the Royal Society argues even 37 is inadequate for multi-decade dunkelflaute.

The bottom line

The honest conclusion isn't 'nuclear is cheapest.' It's that GenCost's nuclear number is highly sensitive to defensible assumptions CSIRO chose conservatively — and that the single biggest lever, first-of-a-kind cost, runs AGAINST nuclear. A mature, government-financed fleet running flat out is competitive. A first Australian plant, with the +120% premium GenCost itself suggests, is not. Both of those are true, and the page shows both.

The full nuclear counterfactual →The Nuclear Case →

Source: CSIRO/AEMO GenCost 2024-25Final Report (Graham, Hayward & Foster, July 2025), App. Tables B.1/B.9 & §2–3; plus submissions from the World Nuclear Association, IEEFA, the Centre for Independent Studies and US EIA. This page critiques GenCost's assumptions — it does not allege faked numbers; GenCost is internally consistent and publishes its full method.