Andrew Hastie called on Australia to dramatically increase investment in artificial intelligence on June 15, 2026, comparing the global competition to develop and control AI to the Cold War nuclear arms race — a comparison that frames AI capability as a strategic necessity, not just an economic opportunity.

Speaking at a forum on national resilience, Hastie argued that Australia needed to reduce its dependence on the United States for AI innovation. The nuclear analogy is deliberate: it suggests that relying on an ally's technology creates a vulnerability, not merely an efficiency problem. During the Cold War, true deterrence required building your own nuclear capacity — borrowed weapons did not provide genuine security. Hastie appears to be making the same structural argument about AI infrastructure: without domestically controlled systems, a nation cannot guarantee access when geopolitical relationships shift.

This argument arrives as Australia's policy community actively debates what "sovereign AI" actually means in practice. The Australian Academy of Technology and Engineering recently released an investment blueprint arguing that government funding is essential to building sovereign AI capability and capturing a larger share of AI's economic value. The academy's language is less militaristic than Hastie's but reaches the same destination: if governments do not invest, private markets will leave national AI capacity dependent on foreign capital, foreign data centres, and foreign decisions about how data is governed.

The strategic logic is direct. A country without meaningful AI infrastructure — the data centres that store information and run computations, the trained AI models, the supply chains for computer chips — cannot ensure that infrastructure will remain available, unchanged, or secure when relationships with supplier countries become strained. The US-Australia alliance has proven durable, but strategic planning does not rest on current political conditions. Hastie, a former defence minister with extensive exposure to intelligence agencies, is working from a standard risk assessment that defence establishments maintain.

Where the nuclear comparison becomes both useful and imprecise is that nuclear weapons had a clear dividing line: you either possessed a working deterrent or you did not. AI capability exists on a spectrum, with interdependencies across every level — from manufacturing the computer chips that power AI systems to the costs of running those systems at scale. Australia cannot realistically build every layer of AI infrastructure at home within any feasible budget. The practical question that Hastie's statement raises is which layers of that infrastructure matter most for sovereign control. Control over computing power, control over data, and control over the AI models themselves each carry different costs and different strategic benefits.

Australia currently invests less in state-directed AI development than most comparable wealthy countries. The ATSE blueprint makes this gap clear. Hastie's remarks at the forum, framed around national security rather than industrial policy, represent an attempt to shift this debate from economic ground to security ground — territory where higher spending is politically easier to defend and where relevant government ministers sit closer to the centre of political authority.

The nuclear analogy also carries an implicit warning about timing. Nuclear programmes that fell behind their competitors rarely caught up. If Hastie's comparison holds weight, the cost of delay is not a missed economic opportunity but permanent strategic subordination. This argument typically persuades defence establishments, and it is clearly Hastie's intended audience.

Whether the government moves from strategic statements to actual funded programmes remains an open question. Discussion of AI sovereignty has occurred in Australian policy circles for years without generating the level of investment needed to close the capability gap the ATSE blueprint identifies. Hastie's intervention may raise how much political attention this issue receives, but political attention and government funding are not the same thing.