For most of the last few years, Australia's energy debate has run along a fairly settled fault line: Labor backing renewables plus storage, the Coalition pushing nuclear, and the argument mostly playing out in election costings and Senate committee rooms. The AI data-centre boom has pulled that fight back into daily politics, because data centres are exactly the kind of customer the grid was never really built for — enormous, constant, inflexible demand, arriving all at once, in a system still trying to work out how to run on wind and sun.
Why data centres changed the conversation
There are currently around 160 data centres operating in Australia, with more than $150 billion in further investment reportedly lined up as AI adoption accelerates. Unlike a household or even most factories, a data centre doesn't ramp its demand up and down with the time of day — it wants power 24 hours a day, every day, at a scale that can rival a small city. That's a difficult customer to plug into a grid that's increasingly built around intermittent generation, and it's forced a question that had mostly stayed theoretical into something with real investment decisions attached: what, exactly, is supposed to back up wind and solar when the sun isn't out and the wind isn't blowing, at the scale AI needs?
What Labor proposed — and what just happened to it
Anthony Albanese's answer, announced in the middle of 2026, was a mandatory standard: new AI data centres would need to deliver as much energy into the grid as they take out, sourced from new renewable projects, batteries, pumped hydro and gas as backup. It was a genuinely ambitious ask, and it didn't survive contact with the states for long. Queensland and the Northern Territory publicly rejected the mandate in late July, and at the National Cabinet meeting on 27 August 2026 — after what was described as "considerable pushback" from both states — the federal government granted them a carve-out allowing data centres in those jurisdictions to run on coal and gas rather than renewables. Queensland Premier David Crisafulli welcomed it as a way to "bring on more power" that was "affordable, reliable and sustainable" while "keeping driving down power prices." The reaction from the left flank was blunt: Greens Senator David Shoebridge called it "a complete collapse by Labor within barely six weeks" of the policy being announced. South Australian Premier Peter Malinauskas tried to hold the line rhetorically, arguing renewables would still "principally" power the sector even as he acknowledged states would differ, while energy academic Gregor Verbic warned that gas as a baseload substitute, rather than storage-firmed renewables, sits awkwardly next to Australia's emissions targets. Whatever position you land on, the speed of the backdown — a national standard announced and substantially walked back within six weeks — is itself a data point about how much political weight federal energy policy can currently carry against a determined state government.
The nuclear moratorium, and why it's still there
Sitting underneath all of this is a federal ban on nuclear power that's now more than a quarter-century old. It wasn't the product of a single deliberate decision so much as a side effect of one: in 1998, the Howard government accepted a Greens amendment to what's now the Australian Radiation Protection and Nuclear Safety Act in order to secure Senate support for a research reactor at Lucas Heights, and that amendment prohibited the development of other nuclear facilities. The Environment Protection and Biodiversity Conservation Act reinforces the ban. Lifting it would require repealing both federal prohibitions and, in several states, separate state-level bans as well — meaning nuclear's path back onto the table runs through multiple parliaments, not just one policy announcement.
The Coalition's case, and the costing fight that followed it
The Coalition took a plan to lift the moratorium and build seven nuclear plants to the 2025 election, costed by Frontier Economics at just over $300 billion for around 14 gigawatts of capacity by 2050, with the first plant not operating until 2036 at the earliest — and claimed, on its own modelling, that this would be $263 billion cheaper than Labor's renewables-based path. The government and CSIRO disputed the comparison sharply: CSIRO's GenCost analysis put nuclear at roughly twice the cost of renewables, and Energy Minister Chris Bowen identified what he called "three fatal errors" in the Coalition's modelling — including an assumption that Australia's future energy needs would be almost 50% lower than Labor's own projection, which mechanically made nuclear look cheaper, and an accounting approach that spread nuclear's costs across 50 years while comparing them against a 25-year costing window. The nuclear plan was a significant factor in the Coalition's 2025 election loss, and Bowen's position since has been unambiguous: "Nuclear is not an answer for Australia any time." Shadow Energy Minister Dan Tehan has kept the case alive from opposition, arguing the moratorium now actively deters AI data-centre investment and pointing to Google and Amazon's nuclear-backed data centre deals overseas as proof the technology is investable at exactly this scale.
Is this actually the right debate to be having?
Grattan Institute's Tony Wood has described the renewed nuclear argument as a "political sideshow" — and there's a straightforward case for that view. Even under the Coalition's own timeline, a first nuclear plant wouldn't be operating until 2036, six years after the 2030 renewables target date this whole discourse is organised around. Nothing about the nuclear debate changes what needs to be built in the next four years: transmission lines, batteries, pumped hydro, and enough gas peaking capacity to cover the gaps while storage scales up. Google and Amazon's overseas nuclear deals are also a different proposition to what the Coalition costed domestically — those are typically agreements with existing or soon-to-be-restarted plants and small modular reactor developers, not a from-scratch fleet of seven new plants in a country that has never operated a commercial nuclear power station.
What's actually worth arguing about tonight
The nuclear question is a legitimate one about Australia's energy mix in the 2040s and beyond. But the data-centre backdown at National Cabinet is a live example of a much more immediate question: when a genuinely ambitious national energy standard meets a state government that doesn't want it, what actually happens — and does "renewables first, with gas as backup" survive contact with real investment decisions, or does gas quietly become the default the moment a state pushes back? That's the version of "nuclear, gas or renewables" that's actually being decided in real time, not in 2036.
Sources
- ABC News — "Renewables, nuclear, gas? Data centre boom energises political debate"
- Bega District News — "PM backs down on renewables mandate for data centres"
- ABC News — "Coalition unveils nuclear costings, assuming smaller economy and higher emissions"
- Energy Council — "Nuclear power for Australia: A potted history"
- Climate Council — "CSIRO confirms nuclear fantasy would cost twice as much as renewables"