Australia is in the middle of a building spree that most people will never see. Behind fenced compounds on the fringes of Sydney, Melbourne and increasingly regional centres, developers are racing to erect the windowless sheds that house the servers powering the artificial intelligence boom. The money involved is enormous, the timelines are compressed, and — according to a growing chorus of economists and planners — the side effects could reach well beyond the technology sector.
A report in The Guardian has crystallised a concern that has been building quietly among macroeconomists: that the sheer speed and scale of AI datacentre construction could add to inflationary pressure at a delicate moment, while soaking up land, energy and skilled labour that the country desperately needs for other things — most pointedly, housing.
The news
The warning is essentially about competition for scarce inputs. Datacentres are capital-intensive, power-hungry projects that demand large parcels of well-located, grid-connected land, thousands of construction workers with specialised skills, and a reliable supply of electricity and water. Each of those is already in tight supply in Australia. When a new source of deep-pocketed demand enters the market, basic economics suggests prices rise — for industrial land, for electricians and data cabling specialists, for grid connections, and ultimately for the energy everyone else consumes.
That is the mechanism experts cited by The Guardian point to when they talk about inflation. It is not that a server hall directly lifts the price of a weekly grocery shop. It is that a torrent of investment concentrated in a short window can bid up the cost of construction and energy inputs across the board, feeding through to the broader price level at precisely the time the Reserve Bank has been trying to coax inflation back inside its target band. Add in the fact that much of the equipment — the graphics processing units, cooling systems and switchgear — is imported and priced in US dollars, and the pressure compounds.
The housing angle is where the story becomes visceral for ordinary Australians. Datacentres and new housing estates want many of the same things: flat, serviced land near infrastructure, and access to power and water connections. In a country that has set itself an ambitious national housing target and is still failing to hit annual completion goals, the prospect of hyperscale campuses hoovering up industrial and peri-urban land, and competing for the same trades who could be framing homes, is politically combustible.
Two views on the risk
Not everyone accepts the alarm. Industry advocates argue that datacentres are a foundational piece of modern economic infrastructure — the plumbing for banking, health records, government services and the AI tools that businesses are increasingly built on. On this view, refusing the investment would simply push it offshore, leaving Australia dependent on capacity in Singapore, Japan or the United States while forfeiting the construction jobs, rates revenue and sovereign-data benefits that come with hosting the facilities locally. Datacentre operators also point out that the land footprints, while large, are modest compared with the total pool of industrial land, and that many projects are sited on parcels unsuitable for housing anyway.
The sceptics counter that the comparison is too neat. The problem is not the total quantum of land but the specific, grid-connected, water-serviced land that is genuinely scarce — and that is exactly what datacentres compete for. They also question the jobs dividend: construction is labour-intensive, but a finished datacentre employs relatively few people to run. A facility that occupies a large block and draws the power of a small city while providing a few dozen ongoing jobs is, critics argue, a poor land-use trade compared with the housing or advanced manufacturing that same site could host. The two camps agree on the facts and disagree sharply on the priorities.
The Australian stakes
For Australia, the stakes are sharpened by timing. The country is simultaneously trying to bring down inflation, build hundreds of thousands of new homes, decarbonise its electricity grid and position itself as a credible player in the global AI economy. Those four objectives are now in direct tension. Every megawatt committed to a datacentre is a megawatt that cannot power homes or heavy industry without new generation being built. Every electrician pulled onto a hyperscale project is one not wiring a housing estate.
Energy is the crux. Australia’s grid is mid-transition, with coal plants retiring and renewables plus storage still being built out. Large, always-on datacentre loads change the demand picture materially, and if new clean generation does not keep pace, the result is either higher wholesale power prices for everyone or continued reliance on gas and coal that undercuts emissions targets. Water is a quieter concern, particularly for evaporative-cooled facilities in drought-prone regions. For state governments courting the investment with incentives and fast-tracked approvals, the challenge is capturing the upside — jobs, sovereign capacity, tax — without exporting the costs onto households already stretched by mortgages and power bills.
There is also a regional dimension worth watching. Some operators are eyeing sites well outside the capital cities, drawn by cheaper land and, in places, proximity to renewable generation. Handled well, that could bring investment and grid upgrades to regional communities. Handled poorly, it risks importing capital-city land and energy pressures into towns with even thinner infrastructure buffers.
What’s next
The policy questions now land squarely with governments. Planning ministers will face pressure to decide whether datacentres should compete for land on equal footing with housing or be steered to designated precincts. Energy regulators and the market operator will need to plan for large new loads without letting them destabilise prices or the decarbonisation timetable. And the Reserve Bank will be watching whether a construction and energy surge adds to the inflation it has fought to contain.
None of this argues for slamming the door on AI infrastructure — the economic and strategic case for hosting it locally is real. But the warning from experts is a call for coordination rather than a free-for-all: to treat land, power, water and skilled labour as the genuinely scarce national resources they are, and to decide deliberately how much of each the AI build-out should be allowed to claim. As the diggers keep moving on the city fringes, that reckoning is arriving faster than the policy to manage it.
Sources: The Guardian.
