What is a peripheral middle power to do?

Some initial thoughts on the 2026 National Defence Strategy.

War is the extension of politics, and it is worth understanding the political context of the 2026 National Defence Strategy (NDS26). This is the document the Labor government is taking into the next election, calibrated—at least in part—to wedge the Coalition on defence. Defence Minister Marles, at the National Press Club launch on 16 April, did not bother to conceal that intent. His spray at independent commentators was less subtle still—and revealing. In a period that Marles himself describes in the Foreword as one of ‘fracture, rivalry and disorder’, that kind of political management instinct is precisely what good strategy cannot afford.

NDS26 is a hinge document—or should be. It arrives now when the assumptions underpinning Australian strategy for eight decades are under structural pressure. A document equal to the moment would test those assumptions explicitly: the value of the alliance under a transactional Washington, the viability of a sovereign industrial base, the logic of deterrence without a nuclear guarantee. NDS26 does not do that. It reflects continuity—some of it path dependency, some convenience, and much of it unexamined assumption. The result is a set of internal tensions the document neither resolves nor acknowledges.

Those tensions include: the contradiction between self-reliance rhetoric and structural dependency; the hollowness of the ‘defence industry’ framing given what Australia actually produces; the mismatch between a biennial production cycle and the timescales on which strategic circumstances are shifting; and, the weight placed on a denial strategy in an environment that is disrupted in ways the denial concept was not designed to handle.

There are also two significant absences—extended deterrence¹ and AI governance²—which, given how much weight extended deterrence bears for Australian security now, and how much is being placed on AI for the future, are not minor omissions. They are structural gaps.

I will address each of these in turn and eventually propose some markers—flags or gates—that will tell us whether Australia’s strategic posture is tracking adequately or whether we are in more serious difficulty than the official narrative allows. I also direct readers to Mick Ryan’s assessment of force structure implications and Marcus Hellyer’s analysis of the funding question; I share their concerns.

The structural condition

Strip NDS26 of its political packaging and what remains is a recognisable strategic problem: a mid-sized polity occupying a strategically significant maritime zone, with a population and industrial base insufficient to generate autonomous hard power, attempting to articulate a security posture during a period when the hegemonic order that made its position viable is losing either the capacity or the will to sustain it.

Australia’s geographic condition—a large, resource-rich, sparsely populated landmass at the junction of two oceans, adjacent to the most demographically and economically dense region on earth, dependent on maritime trade routes it cannot independently secure—has been structurally constant for the last couple of centuries. What has changed is the systemic context in which that condition must be managed.

For seven decades, the post-World War II security architecture offered Australia a structurally elegant solution: sovereignty effectively outsourced to a benevolent hegemon; defence spending optimised for expeditionary contribution rather than independent deterrence; a deliberately shallow industrial base, because deep supply chains were unnecessary when the guarantor had them.

That arrangement was self-reinforcing. The alliance produced security, which validated the alliance, which reduced the pressure to build autonomous capacity, which deepened dependency, which made the alliance more valuable. A clean reinforcing loop—stable, until the conditions that made it stable began to shift.

Australia sits within a hub-and-spoke security architecture with Washington as benevolent hegemon and hub. The Five Eyes, AUKUS, the Quad are not a dense, redundant network of peer relationships. They are spokes that happen to share a hub. The system’s resilience is therefore a function of hub reliability, not network topology. If the hub’s behaviour becomes inconsistent—not necessarily adversarial, simply unreliable—the spokes have no structural substitute. That vulnerability is not addressed in NDS26.

The phase transition risk is not primarily military. The scenarios that would force a qualitative shift in Australian strategic behaviour are: a Washington that conditions alliance commitments on economic or political compliance, converting the alliance into a transactional exchange; a regional incident forcing Australia to choose between ally preference and its own strategic assessment; or a technological surprise rendering current procurement logic obsolete before it is realised. Ukraine and Iran are both relevant here—not as models but as illustrations that threshold crossings happen faster than procurement cycles.

The inflection point is underway. Its full structural consequences have not yet arrived. Australia is in the early-to-middle phase of a hegemonic interregnum — between the clear end of the old order and the consolidation of whatever follows.

The fitness landscape³ is shifting. For seven decades, Australian defence adequacy was measured by contribution to alliance operations, interoperability with US forces, and a credible but non-provocative regional posture. That produced an optimal solution: small, professional, expeditionary, deeply integrated with its main ally. It is no longer optimal—or at least it is not, given the resources government is prepared to commit.

The most dangerous structural dynamic within NDS26 is that the strategy is calibrated to a deteriorating but not yet collapsed environment. If the environment deteriorates faster than the industrial base can be built—which is likely—Australia arrives at the crisis point less capable than the strategy assumed, having made investments that cannot be rapidly converted into immediate capability. The strategy is a slow response to an accelerating problem. That gap will not close within the timeframe NDS26 implicitly assumes.

In a world increasingly shaped by hard power and strongman logic, where the Clausewitzian centres of gravity⁴ are distant from Australian shores, and where Australian dependency extends well beyond Washington to encompass the full range of inputs that make a modern democracy function—the question is not rhetorical. What is a peripheral middle power to do? The answers are not to be found in NDS26.

1. To be precise, extended nuclear deterrence is mentioned once, paragraph 1.17: ‘For Australia, the best protection against the increasing risk of nuclear escalation remains United States’ extended nuclear deterrence on the one hand and the pursuit of new avenues of arms control on the other.’

2. Some may see this as pure legalese or bureaucratise. Far from it. It is critical, and should not be deferred as after-thought. AI will radically change how threats are interpreted and warfare is planned, undertaken and reported. It challenges the foundational tenets of civil-military relations as they have been understood since the mid-twentieth century. For example: ‘Algorithmic deference’ runs contrary to Huntington’s ‘objective control’; the understanding and control of these technologies lies largely in the corporate sector, leading to a blurring of corporate values and incentives and military doctrine—and possibly more disputes as between the Pentagon and Anthropic; and Peter Feaver’s principal-agency framing of civilian control will effectively shift from oversight of personnel to oversight of ‘upstream’ system design.

3. In complex adaptive systems, a fitness landscape is a mathematical representation of the correlation between the physical or behavioural characteristics of an organism, technology, organization, or other construct and its ability to survive. It is a surface that shows how changes in phenotype affect fitness, and it can be used to study the evolution and adaptation of different entities within a system. Systems generally move to higher peaks—levels of fitness—but such peaks may be suboptimal, forcing a shift in shape or behaviour. Landscapes also shift over time, forcing adaptation.

4. Defined, at the level of a state, by Clausewitz as ‘a certain center of gravity develops, the hub of all power and movement, on which everything depends.’ Carl von Clausewitz, “On War,” in Carl von Clausewitz: On War, ed. and trans. Michael Howard and Peter Paret (Princeton: Princeton University Press, 1984), 595–96.