Every fresh wave of artificial intelligence arrives with an awkward companion: a bill, paid in electricity. Training and running large models is one of the most energy-hungry activities in modern computing, and the data centres that house that work are now a live concern for grid operators from Virginia to Singapore. So it was notable this week when IEEE Spectrum pointed its lens at an unlikely candidate for solving the problem: Melbourne.
The premise of the piece is simple but consequential. As the world races to build the compute needed for the AI boom, the question is no longer just where the servers go, but what powers them. Melbourne, sitting inside a state that has committed hard to renewables, is presented as a place where the two curves, surging AI demand and expanding clean generation, might actually be made to bend toward each other rather than pull apart.
Why the question matters now
Globally, the numbers are moving fast. The International Energy Agency has warned that electricity demand from data centres could roughly double by 2030, driven substantially by AI workloads. In several markets that surge has already collided with the energy transition: operators have delayed coal retirements, fired up gas peakers, or in some cases signed deals for new nuclear capacity to keep hyperscale campuses humming. The uncomfortable subtext is that AI, sold as a tool to help decarbonise the economy, risks becoming a fresh source of emissions if it is simply bolted onto a fossil-heavy grid.
That is the tension the Spectrum feature probes. Victoria has legislated an ambitious renewable energy target and is building out offshore wind, large-scale solar and storage. If new AI infrastructure can be timed and sited to draw on that clean supply, rather than forcing more firming from gas, the state could offer a template that other jurisdictions, drowning in demand and short on green electrons, cannot.
The optimistic case
Proponents of the Melbourne story point to a few structural advantages. The city hosts a dense cluster of universities and research institutions, alongside a CSIRO presence, giving it both the talent and the appetite for the kind of systems research that “AI plus energy” demands. It sits within a National Electricity Market that is decarbonising quickly by international standards, with renewables regularly supplying large shares of grid power on sunny, windy days.
There is also a genuine engineering opportunity. AI training, unlike a hospital or a data-hungry bank, can in principle be scheduled. If a chunk of compute can be run when wind and solar are abundant, and eased back when the grid is tight, data centres shift from being a passive load to a flexible one that helps balance a renewable-heavy system. That “demand flexibility” idea is one of the more credible ways to reconcile AI’s appetite with a clean grid, and it is exactly the sort of problem a research-rich city is well placed to work on.
The sceptical case
Not everyone is convinced the alignment is as neat as the framing suggests. The hard truth is that most commercial AI workloads are not flexible today: customers expect chatbots and inference services to respond instantly, around the clock, which means baseload-style demand that does not politely wait for the wind to pick up. Firming that load still leans on gas or imports in many hours of the year.
There is also a water and land dimension that headline energy figures tend to obscure. Large data centres consume significant volumes of water for cooling and compete for industrial land and grid connections, connections that are already a bottleneck for renewable projects trying to plug in. Critics argue that unless planning, transmission and tariff design are handled deliberately, new AI campuses could crowd out the very clean-energy build-out they are meant to complement. Enthusiasm about a city “aligning” AI and clean power, in this view, risks running ahead of the physical constraints of the grid.
What it means for Australia
For an Australian audience, the Melbourne question is more than a curiosity, it is a preview of a national decision. Sovereign compute has become a policy talking point in Canberra, with governments and industry arguing that Australia needs domestic AI infrastructure rather than renting it all from offshore hyperscalers. Every serious version of that ambition runs straight into the energy question: new compute means new load, and where that load lands will shape both electricity prices and emissions.
Australia has a rare card to play here. Unlike much of the northern hemisphere, it has abundant sun, wind and space, and a grid that is decarbonising at pace. If the country can pair AI build-out with renewables, using flexible workloads, co-located storage and smart siting near generation, it could turn a liability into a competitive advantage, marketing “clean compute” to a world short on it. Get it wrong, and AI demand becomes another reason to keep gas running, undercutting climate commitments and raising bills for households already feeling the pinch.
The stakes are concrete for Victorians in particular. The state’s renewable targets, offshore wind zones and storage rollout are already ambitious; layering fast-growing data-centre demand on top will test whether that plan has the headroom regulators assume. It also raises questions the Australian Energy Market Operator and state planners will need to answer soon: how to connect large new loads without stalling renewable projects in the same queue, and how to price flexibility so that data centres have a reason to shift with the grid rather than fight it.
What’s next
The honest answer is that the “alignment” IEEE Spectrum asks about is not yet settled, it is a bet. The technology to make AI workloads grid-friendly exists, and the clean supply is being built, but the commercial incentives, planning rules and market signals that would knit them together are still catching up. Expect the next year to bring more announcements of Australian data-centre projects paired with renewable power-purchase agreements, more scrutiny of their water and grid impacts, and growing pressure on governments to set rules before the concrete is poured.
Whether Melbourne earns the title Spectrum floats will depend less on any single project than on whether Australia treats AI’s energy hunger as a planning problem to be solved rather than a growth story to be celebrated. The city has the ingredients. The question is whether the country turns them into a recipe.
Sources: IEEE Spectrum.
