The country is arguing about defence again, and the argument is about a number. Two and a half per cent of GDP by 2027, an ambition for three per cent in the next parliament, a NATO target of three and a half per cent plus more for infrastructure by the mid-2030s, against a little over two per cent today. Whatever figure wins, the same word sits underneath it doing more work than it can bear. We are arguing about how much to spend before we have agreed what the spending is for.
Defence is treated as one thing the money buys more or less of. It is at least four. There is deterrence, persuading an adversary not to start. There is warfighting, prevailing in the conflict you expect. There is national resilience, absorbing attacks on energy, supply, data, and population that never reach a battlefield. And there is endurance, the industrial and human capacity to regenerate forces and keep producing once a war has begun and the stockpiles are gone. These pull in different directions and reward different spending. A percentage of GDP cannot tell you which of them you are buying.
A share of GDP is the most legible number in the whole debate, which is exactly why it dominates it. It is an input measure: money committed, not safety produced. Lord Case, former Cabinet Secretary, put it plainly on BBC Radio 4’s The World This Weekend (14 June 2026): “We end up talking about how well defended we are as a nation via how much money we spend on it. It’s a really stupid metric.” Like any proxy-capture, once the figure becomes the target it starts to stand in for the thing it was meant to measure, and a government can hit the number while buying the wrong mix. The recent funding settlement made the tension visible. Jack Watling of the Royal United Services Institute, speaking on the same programme, described a department committed to large capital programmes, a new combat aircraft and new submarines for the nuclear deterrent, whose costs already exceed the Treasury’s settlement before any modernisation begins. Protect those programmes, and there is, in his words, “very little headroom to buy drones and to build the industrial base to support those weapon systems that have been shown to be important now in Ukraine.” The savings fall on conventional mass and on people, leaving a force that may produce something useful in 2035 while remaining unmodernised and unready in the short term. That is a real strategic choice. It is not one a budget headline can make for you.
Underneath the figure sits the question the figure lets everyone avoid: what kind of capability actually keeps us safer now. Ukraine has shown cheap, attritable drones destroying platforms that cost thousands of times more, which raises a fair suspicion that some traditional and expensive capability survives less on present effectiveness than on institutional gravity, the sunk doctrine, the prestige of large platforms, the industrial jobs, the procurement cycles measured in decades. Asking whether we keep doing old things partly out of deference to tradition is not unserious. It is close to the central question, and it is being asked far more inside the institution than in the public argument about the percentage.
But the resilience answer is not simply to swap the old and expensive for the new and cheap. Betting the force on drones because the last war rewarded them is the same mistake as betting on tanks because the one before it did: it optimises hard for a single predicted future and calls the result efficiency. As overoptimisation describes, the narrowing looks like prudence right up until the unscheduled thing arrives. Resilience is the capacity to be wrong about which war comes and survive the error, which argues for a spread of capability and, above all, for endurance: the unglamorous ability to replace losses and out-produce the other side. Ukraine’s deeper lesson is less that drones win than that the side which can mobilise and adapt wins, which is a resilience point and not a shopping list. It is the same move from-optimised-to-resilient sets out, away from a single optimised bet and towards connected, regenerating capacity.
So even inside the military, the home ground of the word, resilience is not straightforward. The spending debate flattens four different meanings of safety into one cash figure and then argues about the figure, which is how a question of strategy and value gets laundered into a question of accountancy. The number is not nothing. But a country can spend three per cent of its GDP on the wrong century’s war, and the percentage will not warn it. The argument worth having is about what defence is for. The budget is downstream of that, and we keep starting at the wrong end.
Related
from-optimised-to-resilient : the same move at the level of the whole system, away from a single optimised bet towards connected, regenerating capacity.
resilience-as-abundance : endurance read as productive capacity rather than cost, the diversity that holds also being the diversity that sustains.
overoptimisation : how narrowing to one predicted future reads as efficiency until the unscheduled threat arrives.
proxy-capture : the GDP percentage standing in for the safety it cannot itself measure, and then being managed for its own sake.