Every few years an Australian government unveils a plan for the next wave of technology, and every few years the same worry surfaces: are we building the thing, or just buying it from someone else? That question sits at the centre of a pointed column in the Australian Financial Review, which argues that the Albanese government’s approach to artificial intelligence carries the very flaw that left the country a follower rather than a leader in earlier technology cycles.
The masthead’s headline puts it bluntly, warning that Albanese’s AI plan has the same problem that got Australia into this mess. The “mess”, as the column frames it, is a long habit of treating transformative technology as something to consume and to regulate, rather than something to invent, own and export. On that reading, a national AI framework built mainly around encouraging adoption and setting guardrails risks locking in the same dependence that has defined Australia’s relationship with software, cloud computing and advanced manufacturing.
The context behind the critique
The government has spent much of the past year assembling the scaffolding of an AI agenda. That includes moves towards a coordinating body, often described as an Office of AI, and the promise of a national framework to give business and the public service a clearer sense of the rules. Ministers have leaned heavily on the productivity story, pitching AI as a way to lift sluggish output growth and, by extension, to ease the pressure that keeps interest rates uncomfortably high.
Adoption is where the immediate gains look easiest to bank. If nurses, tradespeople, accountants and public servants use AI tools well, the argument goes, the whole economy becomes more productive without anyone needing to build a frontier model in a Sydney or Melbourne data centre. It is a sensible instinct, and it reflects the reality that Australia is a mid-sized economy competing against companies whose research budgets dwarf the entire local sector.
Two ways of reading the same plan
The AFR’s case is that adoption-first thinking, however pragmatic, quietly concedes the more valuable ground. If the tools are all imported, the intellectual property, the margins and the strategic leverage stay offshore. Australia becomes a very good customer, paying licence fees and hosting other people’s infrastructure, while the enduring economic value accrues in the United States, China and a handful of other places willing to fund the hard, expensive work of building the technology itself. The column ties this to earlier episodes where the country had the talent and the early lead but let the commercial upside slip away.
The counter-view, which the government and plenty of economists would put forward, is that trying to out-build the giants would be a costly vanity. Capital is scarce, energy is constrained and the local market is small. On this argument, the smart play is to specialise: pour resources into the niches where Australia has a genuine edge, such as agriculture, mining, health and climate science, and let the world’s hyperscalers carry the burden of training the largest general-purpose models. Regulation, in that framing, is not timidity but a way of building the public trust that adoption ultimately depends on.
Both positions contain a good deal of truth, which is part of why the debate keeps recurring. The uncomfortable question the AFR raises is whether the government has actually chosen between them, or whether it has defaulted to the path of least resistance because building is harder, slower and politically riskier than convening a taskforce.
What it means for Australia
For all the abstraction, the stakes are concrete. Australia is in the middle of a data-centre construction boom, with billions of dollars flowing into the land, power and cooling that AI workloads demand. That investment is real, and it creates jobs, but hosting the computation is not the same as owning the models that run on it. FluentSea has repeatedly returned to this measurement problem: the difference between an economy that builds AI and one that merely hosts it can be hard to see in the headline investment figures, yet it shapes who captures the long-run value.
There is also a sovereignty dimension that goes beyond economics. If critical services in banking, health and government come to rely on foreign models and foreign infrastructure, the country’s room to set its own rules narrows. Sovereign capability advocates have been making this point loudly, pushing for locally controlled compute, locally trained models tuned to Australian data, and procurement settings that give homegrown providers a fighting chance against incumbents with far deeper pockets. The AFR column effectively lends that camp a mainstream platform, framing the choice not as national pride but as basic economic self-interest.
The energy angle complicates everything. The same data centres that could anchor a domestic AI industry are also straining an electricity grid already juggling the shift to renewables. Any serious plan to build rather than host has to reckon with where the power comes from, at what price, and who else in the economy gets crowded out. That tension is one reason governments find adoption so appealing: it promises productivity gains without forcing a fight over gigawatts.
What happens next
The immediate test is what the national framework actually contains when the detail lands. A document heavy on principles, consultation and risk management would tend to confirm the AFR’s suspicion that Australia is settling for the role of careful consumer. A plan with real money behind local compute, research commercialisation and procurement preferences for Australian capability would suggest the government has heard the criticism and wants to change the pattern.
Business will be watching the incentives most closely, because that is where intent becomes visible. Grants, tax treatment and government contracts reveal far more about priorities than any ministerial speech. If the coming budget cycle rewards firms that build and export AI, the country may yet avoid repeating its history. If it mostly subsidises firms to buy tools from abroad, the AFR’s warning will look less like a provocation and more like a forecast.
None of this makes adoption a mistake. Getting Australian workers comfortable and capable with AI is worthwhile on its own terms. The argument is about balance, and about whether a plan that starts with adoption ever finds the ambition to move beyond it. That is the debate the country needs to have while the decisions are still open.
Sources: Australian Financial Review.
