Before deciding how to power data centres, decide where to build them
National Cabinet spent last Wednesday discussing how Australia’s data centres will be powered. The harder question – where they should be built – was left off the table.
The meeting was meant to lock in mandatory national standards for large data centres covering energy, water and land use. It didn’t. Queensland and the Northern Territory refused to sign up to rules tying new facilities to renewable generation. Going in, the federal government said it would override both governments if they held out. Following the meeting, Energy Minister Chris Bowen said there would be no carve-outs or exceptions to those rules, only a narrow concession allowing existing coal and gas where a jurisdiction could prove that they would be cheaper than renewables.
Legislation isn’t expected until 2027. The investment pipeline is estimated above A$150 billion, and developers are already moving to secure approvals under current rules. Once granted, the geography is fixed for decades and no later standard will move it.
Beneath the argument about technology sits a proposition every government at the table accepts: an industrial load of this scale should not simply arrive on the grid and leave households and businesses to absorb the cost.
Whatever the fuel, additional load must be matched by new supply. While this tells a developer what to fund, it doesn’t ask where funding new power generation is fastest, cheapest or most useful nationally. Standards for energy, water and land use are performance tests applied to a site the developer has already chosen. None of them tests the choice itself.
In July, I argued that Australia lacks a national siting strategy for data centres, and that a meaningful share of it belongs in northern Australia. Last week’s meeting reinforced the point rather than resolving it. State and territory governments negotiated hard over a standard that would apply overwhelmingly to facilities clustered in Sydney and Melbourne, based on current trends.
Siting is also the cheapest route to the outcome the federal government wants. The binding constraint on data centre growth isn’t land or capital. It’s the speed at which new power generation and transmission can be delivered. A data centre can be built considerably faster than the wind or solar capacity and grid connection needed to power it. That mismatch is what drives the price pressure, and it is sharpest where the network is already congested.
Northern Australia inverts the problem. The Northern Territory sits outside the National Electricity Market (NEM); has its own generation, transmission and distribution; and has Beetaloo gas coming online. Queensland retains public ownership across much of its energy supply chain. This creates an opportunity to match new load with new supply without simply passing the cost through to existing households.
For large, energy-intensive loads, an absolutist approach to new gas risks constraining this opportunity. A more practical model would combine renewables with gas-fired power for firming and reliability, supported where necessary by offsets or other abatement. The Northern Territory and Queensland have land, solar, gas and water resources that give them options the southern grid lacks. Their concession should be seen less as a loophole than as a design feature that can be used to attract new load faster and more cheaply without adding to consumer bills.
National standards don’t have to create identical outcomes. A national framework should set the minimum protections and responsibilities and let jurisdictions compete on how they attract investment. That is what last week’s compromise gestured at without quite saying so.
There is a security argument running alongside the economic one. Australia has roughly 290 data centres, heavily concentrated in two cities – a pattern set by commercial history rather than strategy. The Department of Home Affairs is tightening critical-infrastructure obligations across data centres, energy networks and water. Australia has already accepted the logic of geographic diversification for fuel and supply chains. Applying it to computing is overdue: concentration is a resilience problem regardless of what fuels it.
What follows next is modest. When the federal government legislates the mandatory standards next year, it should consider a locational element: a published assessment of network headroom, water availability, time to new generation and strategic exposure, weighed by the regulator alongside cost. The states and territories already hold most of this information. What’s missing is the obligation to use it.
National Cabinet has spent its energy on how these facilities will be powered. The question it has not yet asked is where they should be located. The answer still points north.
About Author
