The CSIRO can't tell you what power will cost

By Caleb Reed · 2026-07-15

What GenCost actually is

GenCost is an annual capital-cost modelling exercise by CSIRO and AEMO that uses levelised cost of electricity (LCOE) and system-levelised cost of electricity (SLCOE) as screening tools for investor decision thresholds, not as a forecast of household electricity prices.verifiedVerified Source: claytonutz.com^[1] That distinction matters, and almost nobody citing the report in public bothers to make it.

GenCost models the minimum price a generator must recover to cover costs under specific boundary conditions. It does not model full system reliability, it does not account for network costs the way a household tariff does, and it was never designed to answer the question politicians keep demanding it answer: will your power bill go up or down?^[1] That is not a flaw in the methodology. The flaw is in the reading.

How the report gets misread in the political fight

When CSIRO and AEMO released GenCost 2025-26 on 15 July 2026, the headline finding was clean and convenient for one side of the debate: renewable energy supported by storage remains the lowest-cost pathway to net zero, with battery technology costs falling and fossil generation costs rising.^[4] Within hours, that sentence had been stripped of its context and weaponised.

Nuclear advocates read the report as politically motivated. Renewables advocates read it as settled science. Neither reading is honest. GenCost's outputs are one input to a policy decision that involves grid architecture, social licence, planning timelines, fuel supply chains and a dozen other variables the report explicitly does not model.^[2] Treating a screening tool as a verdict is lazy politics dressed up as evidence-based policy.

The formal objections and what they actually say

The Centre for Independent Studies put it plainly in its 2 February 2026 submission to the draft report: GenCost should be treated as an input to public debate on energy policy, not a guide for investor or consumer electricity bills.^[2] That is not a fringe view. It is a methodological point that CSIRO's own documentation supports, even if the agency's press releases do not always foreground it.

The World Nuclear Association's objections go further and get into the numbers. In its 13 February 2025 submission on GenCost 2024-25, the WNA challenged the report's assumption that nuclear capacity factors sit between 53% and 89%, arguing those figures reflect market constraints rather than the technology's actual technical availability.^[3] The WNA also attacked the use of a 30-year operational life for nuclear plants in the headline LCOE figures, noting the draft report itself acknowledged a 60-year life would cut the nuclear LCOE by around 10%.^[3]

The WNA further challenged GenCost's capital cost estimate of around A$29,600/kW for nuclear small modular reactors, pointing to the US Energy Information Administration's 2024 nth-of-a-kind figure of US$8,936/kW, approximately A$14,000, which is less than half CSIRO's figure.verifiedVerified Source: world-nuclear.org^[3] Whether or not the WNA's preferred figures are right, the gap between those two capital cost estimates is too large to wave away. It is not a rounding difference. It changes the conclusion.

CSIRO's defence and its revision record

CSIRO Director of Energy Dr Dietmar Tourbier said "Consultation remains central to GenCost, ensuring the report reflects diverse perspectives and the best available evidence to support transparent decision-making across the energy sector."verifiedVerified Source: csiro.au^[4] AEMO Executive General Manager System Design Nicola Falcon said GenCost "continues to be a respected example of collaboration between CSIRO and AEMO, providing trusted, independent insights that support planning for Australia's future electricity system."^[4]

Those are reasonable things to say. CSIRO does consult widely, and it has revised its methodology in response to past submissions, which is exactly what a credible modelling process should do. But revision history cuts both ways. If the methodology has changed before in response to legitimate challenges, the assumption that the current version is beyond challenge is hard to justify. Good science does not close its own consultation loop and declare the debate over.

Tourbier and Falcon did not say GenCost is a forecast of power prices, or that it should be the sole determinant of technology policy. That restraint is appropriate. The problem is that the political actors quoting the report do not apply the same restraint.

The uranium timing and the single-model problem

Six days before GenCost 2025-26 landed, something notable happened in New Delhi. On 9 July 2026, Prime Ministers Albanese and Modi welcomed the signing of an Administrative Arrangement enabling Australian uranium exports to India for peaceful purposes under the 2015 Nuclear Cooperation Agreement.^[5]

Albanese said: "Australia's natural resources are vital for other countries' energy security and stability, and we look forward to becoming a reliable, trusted supplier of uranium to India."^[5] Nobody is accusing anyone of coordination. But Australia opening uranium export channels to a country building nuclear plants, while simultaneously publishing a report finding nuclear uneconomic at home, is a tension that deserves acknowledgement rather than silence.

The deeper problem is not the timing. Any single model, no matter how carefully constructed, carries embedded assumptions that drive its outputs in ways that are not always visible to the people citing it. GenCost's capacity factor range for nuclear, its assumed asset life, its SMR capital cost inputs: each of those is a choice, and choices can be contested. That does not make GenCost wrong. It makes GenCost one view among several that serious policy must weigh. Australia's energy future is too consequential to outsource to a single annual modelling report, regardless of which technology it favours.

Caleb Reed

Caleb Reed covers breaking news and sport for Bushletter. Fast and verb-led, he writes with a news-wire cadence and no patience for PR spin.