For much of the past two years the story told about artificial intelligence in Australia has been one of relentless momentum: billions of dollars pledged for data centres, sovereign AI ambitions spruiked in Canberra, and a queue of listed companies keen to attach their names to the boom. Now one of the country’s most influential business mastheads is asking whether the boom is quite as substantial as the ticker suggests.
In a pointed analysis, the Australian Financial Review argues that Australia’s AI boom is not what it seems. The thrust of the piece is that the visible signs of activity, the cranes, the land deals and the ASX announcements, are outpacing the harder-to-measure things that actually matter: local productivity gains, durable jobs and value that stays in the country.
The context
The backdrop is familiar to anyone who has followed the sector this year. Global hyperscalers and a clutch of domestic operators have announced enormous data centre commitments, with figures running well into the tens of billions of dollars. The land, power and cooling required to host AI workloads has become one of the hottest categories of infrastructure investment in the country, so much so that grid operators and planning authorities are now openly worried about the strain on electricity supply.
At the same time, governments at both federal and state level have leaned into the narrative. Canberra has floated a national framework and the idea of a dedicated office to coordinate AI policy, while the states compete to attract capital with promises of land and connection. The political incentive to be seen as an AI winner is strong, and few politicians want to be caught calling the top of a boom.
The news
The AFR’s contribution is to separate the spectacle from the substance. Its analysis suggests that a large share of the money being celebrated is capital expenditure on physical infrastructure, much of it destined to run models and services owned by overseas companies. The chips are imported, typically from Nvidia. The frontier models are trained abroad. The intellectual property, and the bulk of the recurring revenue, tends to sit offshore. What Australia is buying, on this reading, is the electricity bill and the real estate, not the crown jewels.
That distinction matters because it changes how you score the boom. A data centre is a genuine investment and it does create construction jobs and some ongoing technical roles. But it is a capital-intensive, low-headcount asset. It is not the same as building a domestic AI industry that captures value, exports services and lifts the productivity of the wider economy. The AFR’s argument is essentially that the country risks mistaking the former for the latter.
Two ways to read it
The sceptical view, advanced by the AFR and echoed by a growing number of economists, is that Australia is a price taker in AI. On this account the nation is repeating a familiar pattern: hosting the infrastructure of a foreign-led technology wave while the higher-margin work happens elsewhere. The productivity dividend that was supposed to justify the spending has, so far, been hard to find in the national accounts, and the jobs data has not shown the dramatic shifts that the loudest forecasts predicted.
The optimistic counter, put by much of the industry and by supporters of the sovereign AI push, is that infrastructure comes first and everything else follows. On this view you cannot build a competitive local AI sector without compute on Australian soil, and the current wave of data centre investment is the necessary foundation, not the finished building. Proponents point to the strategic value of keeping sensitive data onshore, to the defence and government use cases that require sovereign capability, and to the early commercial deployments at the big banks and retailers as evidence that real adoption is underway. From this angle, judging the boom by today’s productivity figures is premature.
Both readings can be partly true at once. The infrastructure is real, the productivity payoff is uncertain, and the question is one of timing and distribution rather than whether AI matters at all.
The Australian stakes
For Australia the stakes are concrete and they are not abstract technology-policy debates. Every gigawatt of data centre demand is a claim on an electricity grid that is already mid-transition, and consumers and industrial users will notice if that demand pushes up prices or crowds out other connections. Water, land and planning capacity are finite too, and communities near proposed sites are increasingly asking what they get in return for hosting energy-hungry sheds full of imported hardware.
There is also a sovereignty and skills dimension. If the analysis is right that the valuable layers of the AI stack remain offshore, then Australia’s long-run position depends on whether it can climb up that stack: training and retaining AI talent, funding local model and application companies, and ensuring public money that flows into the sector buys capability rather than just square metreage. The alternative is a boom that inflates asset values and construction pipelines without moving the needle on national productivity, which is the one metric Treasury actually cares about.
The distributional question is sharp as well. Superannuation funds, listed infrastructure players and property owners stand to do well from a data centre build-out regardless of whether the wider economy benefits. That can create a constituency invested in the narrative of the boom continuing, even if the productivity case never fully materialises. Sorting genuine national benefit from private capital gain is exactly the kind of scrutiny the AFR is urging.
What’s next
The near-term test is whether the productivity numbers start to turn. If AI adoption at Australian firms translates into measurable output gains over the next year or two, the optimists will be vindicated and the current spending will look prescient. If it does not, expect the political conversation to shift from attracting investment to questioning its terms, particularly around energy pricing, planning approvals and what governments demand in exchange for supporting projects.
Watch, too, for a sharper policy line on sovereign capability. The debate the AFR is prodding, about whether Australia is building an industry or merely renting out its grid, is likely to define how the next round of AI commitments is judged. For a country that has been here before with other resource and technology waves, the warning is worth taking seriously: booms are easy to announce and much harder to bank.
Sources: Australian Financial Review.
