The Arsenal

Start with a comparison that should be impossible. Australia's GDP today exceeds that of every WW2 belligerent in 1940 – including the United States. It exceeds the combined economic output of all the Axis powers. In 1940, before computers, before reliable diesel engines, before modern healthcare, the US produced nearly 300,000 aircraft. Russia – a country with a terrible climate that is still poor in 2026 – built 160,000. Wealthy, modern Australia assembled 73 F-18 fighters from mostly imported components between 1984 and 1990, and nothing in the 36 years since.

Well and good, the argument goes – we buy better kit than we could build. The problem is what happened to the price of destroying that kit. The five foundational technologies of the electric stack – motors, batteries, power electronics, sensors, and edge compute – have each decosted by roughly 100× over the past 30 years. The guidance electronics that in 1990 required a government munitions program now ship as the cheapest component in a disposable toy. A $2,000 drone can plausibly destroy a $15M tank, a $110M fighter on the ground, or a $3B warship. This is the observed empirical base rate of every live conflict since 2020: Ukraine, Nagorno-Karabakh, the Red Sea, the cartel wars, the Gulf. It is not a speculative threat.

Russia's Black Sea Fleet was significantly larger than the Royal Australian Navy in 2021. It has since been reduced by around a third on most independent assessments – Casey says 45% – by an adversary with no navy at all, using autonomous surface vessels and anti-ship missiles at a cost-exchange ratio on the order of 1:1000. Ukraine produced more than 2 million drones in 2024 and roughly doubled that in 2025, on a pre-war economy a quarter the size of Australia's. Houthi operations have pushed US carrier strike groups into standoff. Russia has lost more than 4,000 tanks – visually confirmed, all causes – with drones implicated in a fifth to a third of recent losses. Australia's entire tank fleet is 59, most already transferred to Ukraine and destroyed. Australia does not have 4,000 spare tanks with which to learn this lesson.

Let's get quantitative. The Hunter-class anti-submarine frigate will spend eight years under construction until first launch in 2032 and will end up costing over $7B – per hull. For reference, the last Nimitz-class nuclear aircraft carrier cost about US$6.2B in 2009 – Casey rounds this to "a carrier, including the planes", which the receipts don't quite support, but the order of magnitude stands: Australia is paying carrier money for a frigate. It buys the complete ground-up development of Falcon 9 and Starlink. Australia is planning to build six, for a program cost north of $45B – about 1.5 Manhattan Projects – and the entire six-hull program yields 192 missile cells. A single US Arleigh Burke Flight III carries 96. Forty-five billion dollars for two destroyers' worth of missile cells, each hull as vulnerable to explosive jet skis as the fleet Ukraine halved without a navy.

And the ships are the visible half of the problem. Adversarial powers now prefer a winking deniability: cable cuts, industrial espionage, cyber operations against Optus-class infrastructure, diaspora policing, information operations, and containers of cheap drones prepositioned by proxies – the modal form of contemporary conflict, calibrated to fall short of a declaration of war. In 1943 Australian commandos infiltrated occupied Singapore in a disguised fishing boat and damaged six ships, once, at enormous risk. Today a competent high-school robotics team could execute the same raid with autonomous boats, and a well-resourced military could do it at 1000× the scale.

The essay audits the order of battle line by line. The compressed version:

None of the alternatives are technically difficult relative to Australian economic capacity. The Soviet Union delivered most of this list in the 1950s at a fraction of Australia's current real GDP per capita, and each program below could run for under $1B a year – a rounding error inside a $60B budget, and less than a tenth of what AUKUS Pillar 1 is projected to consume annually by the late 2030s:

• Sovereign orbital launch. SpaceX developed Falcon 1 in five years for under US$100M; Rocket Lab repeated the feat a decade later. Gilmour Space is already flying hardware and is chronically under-supported. Without sovereign launch, Australia lives under a sky controlled by others.
• Sovereign satellites. SpaceX ships advanced satellites for under $1M a unit; Planet Labs sells sub-metre imagery to anyone with a credit card. The ADF has no sovereign path to independent space-based reconnaissance at all.
• A million drones a month. Ukraine hit 4 million a year on a quarter of Australia's pre-war GDP. Drone-adequate chip fabs — 1980s-class nodes — cost roughly $5B. Australia bulldozed its only functional fab in 2021 to extend the Sydney Metro.
• Layered missile defence. A system that intercepts 50–70% of incoming threats shifts the adversary's planning; it does not need to be perfect to deter. YouTube hobbyists are building rockets a few extra-legal tweaks from the mission.
• Energy independence. Two refineries is not enough, as the current Gulf crisis is demonstrating on Australian farms. Australia invented the modern solar module, paid China to take it, and could still become the place the world smelts its metals.
• Submarines. The US Navy built the first nuclear submarine powerplant in 1,173 days for roughly $2.5B in current dollars — including the reactor supply chain, from scratch, for the first time in history. Seven decades of design heritage later, Australia is told a foreign-built equivalent costs $30–40B per boat and 15 years. Why do the error bars on this purchase exceed the real cost of the pioneers doing it first?
• AI participation. The near future runs on CCP AI or US AI. Australia doesn't need a sovereign frontier model; it needs to contribute enough of the stack to stay in the conversation, and it is geographically ideal for solar-powered datacentres.

Then there is the recommendation the essay itself labels most politically fraught: nuclear weapons readiness. Casey's argument runs in three legs. Technical: the 1964 Nth Country Experiment showed three physics PhDs with no classified access could produce a workable design in 2.5 years — the physics is not the obstacle. Strategic: the Budapest Memorandum, under which Ukraine traded its arsenal for security "guarantees", is now the canonical demonstration of what those guarantees are worth to a middle power. Political: Australia has no enrichment capability, no test infrastructure, and no coalition willing to sustain the investment — and this, the essay argues, is the only binding constraint, reversible on a sub-decade timescale, so the preparation (civil nuclear industry, workforce, enrichment-adjacent capability) should begin now regardless. That is the essay's position, stated precisely Casey asserts; the treaty exit it would require is its own conversation.

The strategy documents get the diagnosis right. The Strategy of Denial is the correct doctrine for a middle power; deterrence-by-punishment is beyond Australia's resources. The capability program then funds almost none of it: 90% of acquisition spending flows offshore into high-cost, low-count, crewed platforms, in a world where wars are once again decided by materiel production capacity — except now the materiel is cheap, autonomous, and produced by the million. The constraint has never been talent, capital, or technology. Expat Australians work at every frontier technology company on Earth, and an Australian hobbyist holds the world drone speed record. Australia is not undefendable in principle. It is undefendable against the threat model its own documents describe, using the capabilities its own documents fund — and the correction has to start now, not next biennial cycle.