Australia’s construction industry is heading into a split-screen couple of years. On one side, the pipeline of new homes is tipped to thin out through 2027 as higher interest rates and softening house prices sap the appetite to build. On the other, a wave of data centre projects, driven in no small part by the race to stand up artificial intelligence capacity, is propping up the non-residential side of the ledger and changing where the industry’s growth is coming from.
That is the picture painted by new analysis of building approvals, which TechDay Australia reports points to a weaker housing outlook even as big-ticket infrastructure holds up the headline numbers. Building approvals are one of the earliest signals the market gives off, because a project generally has to be approved long before a crane appears on site. When approvals soften, the slowdown in actual activity tends to follow roughly a year or two later, which is why a dip now feeds directly into a gloomier 2027.
The news: a weaker housing pipeline, a stronger data centre one
The core of the analysis is straightforward. Residential approvals, and especially approvals for new houses, are expected to come under pressure as the cost of borrowing stays elevated and price growth cools in parts of the country. For developers and prospective owner-builders alike, the maths on a new dwelling looks harder than it did a couple of years ago, and that hesitation shows up first in the approval statistics rather than in completed homes.
Non-residential building, by contrast, is being lifted by a category that barely rated a mention a decade ago. Data centres are now large enough, and numerous enough, to move the aggregate figures. These are capital-heavy builds, often running into the hundreds of millions of dollars each, and a single campus can dwarf a suburb’s worth of housing approvals in dollar terms. The current boom is being fuelled by demand for cloud computing and, increasingly, by the enormous compute requirements of training and running AI models.
The result is a construction sector whose aggregate numbers can look healthier than the housing story alone would suggest. That matters for anyone reading the top-line approvals data without digging into the composition, because a handful of very large non-residential projects can mask genuine weakness in the homes that households actually need.
Two ways to read it
One view treats the data centre surge as a welcome offset. If residential activity is going to soften regardless of what the industry does, then a pipeline of large, well-funded infrastructure projects keeps trades employed, keeps suppliers busy and keeps the sector from falling off a cliff. On this reading, the AI build-out is doing the economy a quiet favour by smoothing out what would otherwise be a sharper downturn in construction activity.
The more cautious view is that the two trends are not really substitutes for one another. A data centre does not house a family, and the skills, materials and locations involved only partly overlap with those needed for homebuilding. If anything, a construction boom in hyperscale facilities competing for the same electricians, concreters and grid connections could make housing marginally harder and more expensive to deliver, not easier. On that reading, a strong non-residential number sitting on top of a weak residential one is not reassurance so much as a warning that the industry’s capacity is being pulled towards infrastructure at precisely the moment the country says it wants more homes.
What it means for Australia
The timing is awkward. Australia has set itself a national target to build 1.2 million new, well-located homes over five years under the National Housing Accord, and almost every credible assessment already has the country tracking behind that goal. A forecast that points to softer housing approvals through 2027 is, in effect, a forecast that the gap between ambition and delivery could widen rather than close. Governments at both the federal and state level have staked a good deal of political capital on housing supply, so a weaker pipeline is as much a policy problem as an economic one.
At the same time, the data centre boom speaks to a genuine national opportunity. Australia has been positioning itself as a regional hub for cloud and AI infrastructure, and the capital flowing into these facilities represents real investment, real jobs during construction and real digital capacity once the sites are running. The catch is that data centres are voracious consumers of two things Australia is already short of: reliable power and skilled trades. FluentSea has reported on warnings of a looming shortage of tens of thousands of tradespeople needed to service the energy and AI build-out, and the tension between powering data halls and keeping the lights on affordably for households is becoming one of the defining infrastructure debates of the decade.
There is also a geographic dimension. Housing demand is concentrated in the growth corridors of Sydney, Melbourne, Brisbane and the south-east Queensland arc, while data centres cluster where land, power and connectivity line up, which increasingly includes regional sites and purpose-built campuses in Western Australia and elsewhere. The two pipelines are not competing for exactly the same patch of dirt, but they are competing for the same national pool of construction capacity, financing and grid capacity, and that competition only intensifies as AI demand climbs.
What’s next
The near-term signal to watch is the monthly building approvals series, and in particular the split between houses, higher-density dwellings and non-residential work. If residential approvals keep drifting lower while data centre and infrastructure approvals hold the aggregate up, the 2027 outlook described in this analysis becomes harder to dismiss as a blip. Interest rate settings will be the swing factor: any easing in borrowing costs would do more to revive housing approvals than almost any other single lever, which puts the Reserve Bank’s path squarely in the frame.
For the AI and infrastructure sector specifically, the question is whether the data centre pipeline can keep expanding without running into hard constraints on power and labour. Grid connection queues, electricity prices and the availability of skilled trades will all shape how much of the announced capacity actually gets built, and how quickly. For now, the clearest message from the approvals data is that Australia is building a great deal, just not necessarily the homes it most urgently needs, and that the digital infrastructure soaking up so much of the sector’s effort is both a genuine prize and a fresh source of strain.
Sources: TechDay Australia.
