For most of the past decade, the biggest cheque Australia has been asked to write for its security has had the word “submarine” on it. A provocative new column in The Canberra Times wants to reframe that spending debate, arguing that the roughly $11 billion Australia is committing to lift the American and British submarine industrial base under AUKUS represents a strategic misallocation. The money, the column contends, would deliver far more enduring national power if it were poured into sovereign artificial intelligence instead.
The context
The number at the centre of the piece is not plucked from the air. Under the AUKUS pact, Australia agreed to make direct financial contributions to the submarine shipyards of its partners, a commitment that runs to several billion US dollars across the back half of this decade, on top of the eye-watering lifetime cost of the nuclear-powered boats themselves. Successive estimates have put the full Virginia-class and future SSN-AUKUS program somewhere between $268 billion and $368 billion out to the 2050s. Against that backdrop, an $11 billion line item can look almost modest, which is precisely why the column uses it as a lever: if Australia can find that sum to underwrite factories on the other side of the world, why can it not find something comparable to build compute, chips, data centres and AI talent at home?
The timing sharpens the point. Australia has spent 2026 watching a wave of capital chase local AI infrastructure, from Firmus and its multibillion-dollar “AI factory” ambitions to the scramble by banks such as Westpac and NAB to stand up their own frontier capability. Sovereign AI has moved from a niche policy phrase to a mainstream economic argument, and the column lands squarely in that shift.
The argument
The case the column makes is essentially about the nature of modern power. Submarines are a twentieth-century expression of deterrence, expensive, slow to arrive, and dependent on foreign supply chains that Australia does not control. Artificial intelligence, by contrast, is framed as the connective tissue of every future capability, military and civilian alike, from logistics and cyber defence to biosecurity, energy management and the productivity of the wider economy. Spend on a submarine and you get a submarine. Spend on sovereign AI, the argument runs, and you get an input that compounds across defence, industry and government at once.
There is also a sovereignty thread running through the piece. A recurring anxiety in Australian technology policy is that the country rents its most important digital capability from a handful of American hyperscalers, leaving critical decisions about data, models and access hostage to commercial terms and foreign law. Building domestic capability, on this view, is not a nice-to-have but a strategic hedge, in the same category as fuel security or domestic manufacturing.
The other side of the ledger
The counterargument is not hard to mount, and any honest reading of the debate has to give it room. Defence planners would say the comparison is a false choice. Submarines exist to solve a specific problem, the ability to hold an adversary’s shipping and forces at risk across the vast approaches to the continent, and no amount of compute substitutes for a boat that can stay submerged in contested waters for weeks. The AUKUS contributions are also the price of admission to a technology-sharing arrangement that, in theory, gives Australia access to precisely the kind of advanced capability, including AI-enabled systems, that it could not develop alone.
There is a fiscal reality too. The submarine money is largely capital tied to a treaty and a decades-long acquisition schedule. AI spending, by contrast, is a fast-moving operating cost in a field where today’s cutting-edge cluster is tomorrow’s write-down. Critics of the “just build sovereign AI” position point out that pouring billions into data centres does not by itself create sovereign capability if the chips, the model weights and the top researchers still come from overseas. Money is necessary, but it is nowhere near sufficient, and Australia’s thin pool of frontier AI talent is a constraint no budget line fixes overnight.
What it means for Australia
Strip away the rhetorical framing and the column is really asking a question every middle power now faces: in an era when compute is becoming as strategic as steel, how should a country with finite money balance hard defence against digital sovereignty? For Australia the stakes are unusually concrete. The nation is simultaneously trying to fund the most expensive defence acquisition in its history, attract private capital into local AI infrastructure, and avoid becoming a permanent tenant of offshore cloud providers. Those goals compete for the same pool of skilled workers, the same grid capacity, and ultimately the same taxpayer dollar.
The productivity dimension is where the argument bites hardest at home. Modelling circulated this year, including work from EY-Parthenon suggesting AI could touch a third of Australian jobs, has put economic transformation at the centre of the national conversation. If AI is genuinely the general-purpose technology its boosters claim, then under-investing in domestic capability carries an opportunity cost that never appears on a defence balance sheet, but shows up instead as slower growth, weaker firms and a widening gap with peers that moved first. That is the uncomfortable point the column is pressing.
What’s next
Nothing about the submarine program is likely to change on the strength of a newspaper column, and the political consensus behind AUKUS remains firm across both major parties. But the piece reflects a genuine and growing pressure inside the policy debate, one that will only intensify as the federal government works through its promised national AI plan and as more of the budget is claimed by long-tail defence commitments. Watch for the argument to migrate from the opinion pages into estimates hearings and industry submissions, where sovereign AI advocates will increasingly frame compute and capability as a national security asset rather than an economic luxury.
The honest conclusion is that Australia probably needs both, and the real challenge is sequencing and scale rather than a binary choice between a submarine and a server rack. The value of the column is less in its arithmetic than in its provocation: it forces a country that has grown comfortable measuring security in tonnes of steel to ask whether the next threat, and the next advantage, will be measured in teraflops.
Sources: The Canberra Times.
