The Failure of Nuclear Deterrence and the Return of World War

By Matthew Parish

Monday 14 September 2026

For most of the period since 1945, one proposition has stood at the centre of international security policy: nuclear weapons prevent great wars. They may be morally appalling, ruinously expensive and potentially capable of extinguishing civilisation, but precisely because they are so terrible nobody will dare start a major war between nuclear powers. The atomic bomb therefore acquired a peculiar reputation as simultaneously the most destructive and the most peace-preserving invention in human history. Nuclear weapons could destroy the world and, according to the theory, that was precisely why they would save it. Yet there is an increasingly uncomfortable problem with this proposition: the world is once again becoming accustomed to war.

Russia, one of the world’s two greatest nuclear powers, has been fighting the largest European war since 1945. Ukraine’s Western supporters have supplied her with increasingly sophisticated weapons, intelligence and financial assistance while Moscow has periodically threatened nuclear escalation. Nuclear-armed states confront one another elsewhere, directly or indirectly. India and Pakistan remain trapped in an enduring strategic rivalry punctuated by military crises. Israel’s undeclared nuclear status hangs over the perpetually unstable Middle East. North Korea continues to develop nuclear weapons and the missiles with which to deliver them. China is enlarging her arsenal while military competition around Taiwan intensifies. Meanwhile the elaborate system of arms-control agreements constructed during and after the Cold War has progressively deteriorated.

The paradox deserves careful examination because it is too simplistic to say that nuclear deterrence has failed altogether. Nuclear weapons may remain remarkably effective at preventing one very specific event: an intentional, unlimited strategic attack by one nuclear power upon another. Yet preventing such an attack is not the same thing as preventing war. Indeed nuclear weapons may increasingly be having a contrary effect, creating protected spaces beneath the nuclear threshold within which states believe they can wage conventional, proxy, hybrid and limited wars because their nuclear arsenals discourage their enemies from responding decisively. This is a much less reassuring world than the architects of nuclear deterrence imagined.

What deterrence was supposed to achieve

The classical theory of nuclear deterrence was brutally simple. Once two adversaries possessed survivable nuclear arsenals, neither could destroy the other without itself being destroyed. Even a successful first strike would leave enough submarines, mobile missiles, bombers or concealed weapons intact for retaliation. Hence arose the macabre expression “mutually assured destruction”. The logic was not that politicians became peaceful, virtuous or sensible. It was that a certain category of war became irrational because the price of victory could no longer meaningfully be distinguished from the price of defeat.

During the Cold War this appeared, superficially at least, to work. The United States and the Soviet Union competed almost everywhere. They armed proxies, overthrew governments, supported insurgencies, fought wars through allies and accumulated extraordinary quantities of conventional and nuclear weapons. Yet American and Soviet armies never fought a general war against one another. Europe, which had destroyed itself twice within thirty years during the first half of the twentieth century, experienced an extraordinary period without a general continental war between the principal military powers.

Nevertheless this comforting account understates how frequently catastrophe was avoided through luck, individual judgment and improvised restraint. The Cuban Missile Crisis remains merely the most famous example of a much broader phenomenon. Nuclear deterrence was never an automatic machine generating peace. It was a psychological relationship between fallible human beings operating imperfect technologies while attempting to guess one another’s intentions. Deterrence exists only in people’s minds: a nuclear weapon sitting in a silo does not deter anybody by itself. Deterrence requires an adversary to believe that the weapon might actually be used and this, in turn, requires credible political resolve, functioning command systems, comprehensible doctrines and sufficiently rational decision-making on both sides. Every component of that chain can fail.

The Ukrainian lesson

The war in Ukraine illustrates the limitations of nuclear deterrence with unusual clarity. Russia’s nuclear arsenal did not prevent her from undertaking a vast conventional war immediately adjacent to NATO territory. NATO’s collective nuclear arsenal did not prevent Russia from invading Ukraine. Instead nuclear weapons have influenced the boundaries within which the war has been conducted, shaping calculations about which weapons may be supplied, where they may be used and what forms of direct Western involvement might be considered excessively escalatory.

Moscow has repeatedly invoked nuclear escalation, sometimes explicitly and sometimes obliquely. Western governments have consequently had to consider whether particular weapons deliveries, targeting permissions or military deployments might provoke an escalation that neither side ultimately wants. The result is not nuclear peace but nuclear-bounded conventional war. Russia can calculate that NATO will be reluctant to intervene directly because combat between NATO and Russian forces carries some probability, however small, of escalation towards nuclear weapons. Nuclear weapons therefore provide Russia with a degree of strategic protection while she wages a conventional war.

This produces what strategists have long called the stability-instability paradox. Strategic nuclear stability can generate instability at lower levels of conflict. If both sides believe that neither will cross the ultimate threshold, they may become more willing to take risks beneath it. Nuclear weapons can therefore deter Armageddon while enabling aggression, and the distinction between these two effects is becoming increasingly important to understanding contemporary warfare.

The proliferation of thresholds

Modern warfare makes this problem substantially worse because there is no longer a simple distinction between peace and war. A state can sabotage underwater cables, interfere with satellites, conduct cyber attacks, assassinate opponents abroad, manipulate elections, launch drones through proxies, jam navigation systems, attack commercial shipping, conduct influence operations or employ ostensibly private military organisations. None of these activities necessarily resembles the invasion of Poland in September 1939. Nevertheless their cumulative effects may amount to a sustained campaign of warfare by unconventional means.

Nuclear deterrence provides surprisingly little assistance with these problems because nobody seriously imagines launching an intercontinental ballistic missile in response to sabotage of a fibre-optic cable or a cyber attack upon a railway network. The more disproportionate the threatened response, the less credible the deterrent becomes. Nuclear powers therefore discover enormous territories of coercion below the nuclear threshold in which they can compete aggressively without necessarily provoking an existential military response.

The same problem applies to conventional warfare. Would the United States initiate nuclear war over a small island in the western Pacific? Would Britain launch Trident because Russian forces occupied a few kilometres of NATO territory? Would France risk Paris for Tallinn? The purpose of extended nuclear deterrence is to ensure that the answers to questions of this kind remain uncertain, because uncertainty itself contributes to deterrence. Yet uncertainty works in both directions. The defender worries about escalation just as much as the aggressor does and an aggressor may consequently conclude that the defender’s nuclear arsenal restricts her freedom of action rather than strengthening it.

Nuclear weapons have not become obsolete

None of this means that nuclear weapons are useless and it would be dangerous to draw such a conclusion. They remain extraordinarily effective instruments for guaranteeing the survival of certain regimes. Saddam Hussein and Muammar Gaddafi did not possess operational nuclear arsenals when Western military power destroyed their governments. North Korea does possess nuclear weapons and military action designed to overthrow her government is correspondingly almost inconceivable. The lesson has hardly been lost upon governments contemplating their own vulnerability.

Nor have nuclear-armed governments concluded that these weapons no longer matter. Quite the reverse: the principal nuclear powers are modernising their arsenals, developing new delivery systems and reconsidering doctrines that many people imagined would gradually become irrelevant after the end of the Cold War. This creates another paradox. The less nuclear weapons appear capable of preventing conventional wars, the more governments may conclude that they need nuclear weapons to protect themselves during conventional wars. That is a potentially powerful recipe for proliferation.

Ukraine’s history is particularly damaging to the political case for non-proliferation. Whatever the complicated legal and technical realities surrounding the nuclear weapons inherited from the Soviet Union, the political story told around the world is much simpler. Ukraine relinquished the nuclear weapons situated upon her territory, received security assurances and was subsequently invaded by a nuclear-armed neighbour. Other governments, particularly those that fear larger neighbours, will inevitably draw their own conclusions from this sequence of events.

The disappearance of the old nuclear priesthood

There is another reason for concern. The Cold War eventually produced a peculiar professional culture of nuclear restraint. American and Soviet officials understood that their adversaries were enemies but also understood that certain forms of communication had to survive precisely because they were enemies. Hotlines, arms-control negotiations, inspection mechanisms, military-to-military contacts and carefully developed doctrines created a language through which nuclear adversaries could interpret one another. The people who managed nuclear policy became a strange international priesthood whose profession consisted of contemplating the end of civilisation in sufficient detail to prevent it.

Much of that architecture has deteriorated. Arms-control agreements have disappeared or weakened, strategic dialogue has become irregular and political relations between several nuclear powers have become increasingly hostile. One need not romanticise the Cold War to recognise that adversaries then often devoted extraordinary resources to understanding how their opponents thought about escalation. Contemporary political discourse, by contrast, increasingly rewards theatrical threats, public humiliation and maximalist declarations intended for domestic audiences.

This matters because deterrence requires communication. If one state does not understand another state’s red lines, it cannot reliably avoid them. If political leaders speak principally for domestic audiences, threats may be misunderstood abroad. If governments deliberately cultivate ambiguity, adversaries may interpret bluff as preparation or preparation as bluff. The terrifying feature of nuclear warfare is therefore not necessarily that somebody decides rationally to destroy civilisation. It is that governments can climb an escalatory ladder without realising how close they have come to its final rung.

Technology makes everything faster

The Cold War nuclear confrontation was technologically primitive by contemporary standards. Today’s strategic environment contains cyber warfare, autonomous systems, artificial intelligence, hypersonic weapons, anti-satellite systems, sophisticated drones and global communications networks transmitting information almost instantaneously. These technologies have the potential to shorten decision times precisely when nuclear decision-making requires patience, independent verification and human judgment.

Suppose an early-warning system reports incoming missiles. Are they nuclear or conventional? Are they decoys? Is the information genuine? Has the system been hacked? Has an artificial intelligence system misclassified something? Is the attack limited or the opening stage of a disarming first strike? A political leader might have only minutes to distinguish between these possibilities, while knowing that waiting too long might destroy the forces required for retaliation.

The more sophisticated military technology becomes, the more tempting it is to imagine that computers will eliminate uncertainty. In practice they may merely accelerate it. Algorithms can process information vastly faster than people but rapid processing does not transform ambiguous information into certainty. The danger is particularly acute when several technological systems interact in ways nobody completely understands. Nuclear deterrence was developed in an era in which human beings remained visibly at the centre of strategic decision-making. The emerging military world increasingly incorporates machines into detection, targeting, communications and operational analysis and the nuclear powers must be exceptionally careful that automation does not gradually migrate towards decisions that should remain irreducibly human.

How world wars actually begin

There is a persistent misconception that a Third World War would begin with somebody announcing one. The first two world wars demonstrate the opposite. Global wars tend to emerge from the accumulation and internationalisation of crises that initially appear capable of containment.

The First World War emerged from a regional crisis in the Balkans. Austria-Hungary sought to punish Serbia, Russia mobilised, Germany acted in support of Austria-Hungary, France stood behind Russia, Germany invaded Belgium and Britain entered the conflict. A dispute following an assassination in Sarajevo became an industrial war spanning continents. The Second World War was more deliberately expansionist but even it accumulated through successive crises in Manchuria, Ethiopia, the Rhineland, Austria, Czechoslovakia and Poland before the European and Asian wars eventually merged into a genuinely global struggle.

A third world war might similarly be recognised only retrospectively. Future historians might disagree about its starting date. Was it Russia’s invasion of Ukraine? A confrontation over Taiwan? A Middle Eastern war? A clash on the Korean peninsula? A cyber attack upon NATO infrastructure? An incident between India and Pakistan? Perhaps several apparently unrelated wars would gradually merge as alliances, weapons supplies, economic sanctions and military commitments connected them. Contemporary geopolitics increasingly contains precisely these overlapping systems of conflict and that should make us wary of assuming that world war must arrive in a single dramatic moment.

Nuclear weapons may make the opening stages easier

Here lies the darkest paradox. Nuclear weapons may make states more confident that limited wars will remain limited. Russia can tell herself that NATO will not intervene directly because NATO fears nuclear war, while NATO can tell herself that Russia will not attack NATO territory because Russia fears precisely the same thing. China might conclude that the United States would hesitate over Taiwan for fear of escalation while Washington might conclude that Beijing would avoid attacking American bases for identical reasons. Each proposition considered individually may be rational, yet taken together they may create a profoundly unstable international system.

Because each government assumes the other will ultimately retreat from the nuclear threshold, everybody may become more adventurous beneath it. Eventually somebody may discover that an adversary’s red line was genuine. At that moment deterrence has not prevented the crisis but helped to create the conditions in which it arose.

The most dangerous international systems are not necessarily those in which leaders want enormous wars. They may instead be systems in which everybody believes enormous wars are impossible. European leaders in 1914 did not generally expect four years of industrial slaughter and their confidence that escalation could be controlled contributed to making escalation possible. Our nuclear age contains an equivalent temptation: the belief that because global war would be suicidal, nobody will permit one to happen. That proposition is logically attractive but historically reckless.

The return of great-power war

There are already signs that governments recognise the change. Military expenditure is rising, European states are reconsidering conscription, ammunition production, air defence and civil preparedness, China continues to expand her military capabilities, Russia has reorganised significant parts of her economy around war and the United States faces increasingly difficult decisions about distributing military resources between Europe, Asia and the Middle East. Nuclear modernisation is accelerating alongside conventional rearmament rather than replacing it.

This does not mean that a Third World War is inevitable. Predictions of inevitable war are usually irresponsible because political decisions matter, diplomacy matters, military preparedness matters, economic relationships matter and individual leaders matter. Nevertheless the possibility of world war has returned to serious political discussion because the structural conditions that once appeared to make it fantastically improbable are weakening. The world is becoming increasingly multipolar, arms-control mechanisms are deteriorating, regional wars are multiplying, great powers increasingly perceive international politics as zero-sum competition and new military technologies compress decision-making times.

Most dangerously, nuclear weapons have ceased to look like instruments that prevent conflict altogether. Instead they increasingly resemble enormous strategic umbrellas beneath which other forms of conflict can flourish. This does not render nuclear weapons irrelevant; rather it suggests that we misunderstood what they could accomplish. They may prevent certain kinds of war while simultaneously changing the incentives governing others.

Deterrence needs conventional strength

The lesson is not that Western countries should abolish their nuclear weapons unilaterally. In the present international environment that might make matters considerably worse. Nuclear deterrence must instead be understood for what it is: the final layer of national security rather than a substitute for all the others.

A credible defence system requires conventional military forces capable of defeating conventional aggression without immediately confronting political leaders with the choice between surrender and nuclear war. It requires resilient infrastructure, effective intelligence, cyber security, air defence, industrial capacity and alliances whose commitments are politically believable. If nuclear weapons are the only convincing answer to conventional aggression, then the defender may discover that the answer is too terrible to use and therefore too incredible to deter.

Above all, security requires diplomacy. This may sound paradoxical in an era in which international politics is increasingly described as a contest between enemies who have nothing useful to say to one another. Yet hostile governments have the greatest need to communicate precisely because they are hostile. Friendly governments do not require elaborate mechanisms to prevent accidental wars. Adversaries do. The purpose of diplomacy between enemies is not necessarily reconciliation or friendship; sometimes it is merely ensuring that each side understands what the other side might do next. That modest achievement can save millions of lives.

The most dangerous illusion

Nuclear weapons probably did contribute substantially to preventing direct war between the superpowers during the Cold War and we should not casually dismiss that achievement. A conventional Third World War between NATO and the Warsaw Pact might itself have killed tens of millions of people and destroyed much of Europe. Yet from this historical observation arose an excessively broad conclusion: that nuclear weapons had somehow abolished great-power war. They did not. They altered its incentives, pushed conflict into different forms and made certain kinds of aggression impossibly dangerous while rendering others potentially more attractive.

For more than eighty years there has been no second use of nuclear weapons in warfare after Hiroshima and Nagasaki. This is an extraordinary achievement, but it is not proof that there never will be one. The greatest danger facing the contemporary international system may therefore be complacency: the assumption that because nuclear weapons make a Third World War irrational, rationality will invariably prevent one. History offers very little support for the proposition that governments always behave rationally, still less that several governments confronting one another simultaneously will interpret risks in identical ways.

A Third World War is unlikely to begin with hundreds of strategic nuclear missiles suddenly crossing the northern hemisphere. It is more likely to begin with something that initially appears manageable: a regional invasion, a naval confrontation, an attack upon infrastructure, a misidentified missile, a proxy conflict or a military response intended merely to demonstrate resolve. Governments will impose sanctions, send weapons, mobilise forces and issue warnings. Alliances will harden, accidents will occur and each escalation will be presented as a necessary response to the escalation that preceded it. At every stage the participants may continue believing that they remain in control.

Every government involved may insist that it does not want a world war and every government may sincerely believe what it says. Nuclear deterrence will continue hovering above them, discouraging the ultimate act while simultaneously constraining the responses available to each side below it. The danger is that the participants may continue testing one another precisely because they believe the nuclear threshold guarantees that somebody will eventually retreat.

For decades we have regarded nuclear weapons as the reason why there cannot be another world war. We may eventually discover that this was the wrong lesson. Nuclear weapons have made one particular form of world war almost unimaginably terrible, but they have not abolished rivalry, territorial ambition, ideological conflict, miscalculation, technological failure or human folly. They may even have encouraged governments to believe that these ancient forces can safely be indulged because the ultimate catastrophe remains unthinkable.

The most frightening possibility is therefore not that somebody deliberately decides to start the Third World War. It is that nobody does. The world may instead stumble towards one through a succession of individually rational decisions, limited military actions and supposedly controlled escalations, with nuclear weapons looming over every stage and persuading each participant that the others will ultimately give way. By the time that assumption is proved wrong, the question of whether nuclear deterrence prevented the Third World War may have acquired a terrible and unexpected answer.