Editorial Disclaimer: This article is an analytical examination of nuclear deterrence based on historical patterns, publicly available data, and peer-reviewed security research. It does not advocate violence, nuclear weapon use, or proliferation. All statistics cited are drawn from the SIPRI Yearbook 2026 and the Federation of American Scientists.
Last Updated June 2026 | 14 minutes read
9 nations. 12,000+ warheads. No nuclear war in 81 years. The brutal logic behind the number that runs the world.
Image Credit: Leonardo AI
Nine countries hold 12,241 nuclear warheads between them. Not one has been fired in conflict since Nagasaki on 9 August 1945. That is not luck. That is the most uncomfortable security arrangement in human history, and it is the reason the world has not had a direct war between major powers for eight consecutive decades.
Nuclear weapons were built to destroy cities. Their actual function has been to make sure no city is ever destroyed by one again. The logic is precise: attack a nuclear-armed state and absorb retaliation so catastrophic that no military planner on earth can survive or manage it. The rational choice, every time, is not to attack.
That logic has held without a single direct failure between nuclear-armed states since 1945.
Nuclear deterrence functions as a strategic safeguard by making full-scale interstate invasion prohibitively costly. The empirical record since 1945, verified through open-source research by SIPRI, consistently supports the conclusion that credible nuclear deterrence limits direct armed conflict between nuclear-armed states. For the current policy context, see our analysis of nuclear deterrence in 2026 with no treaty and no binding rules.
The Full Story: What Nuclear Deterrence Actually Does
Every military confrontation begins as a calculation. The state considering aggression weighs territorial gain, military advantage, economic benefit, and the anticipated cost of enemy retaliation. For centuries, that final variable operated at a proportional scale. Conventional armies could be countered, territory recaptured, and wars eventually concluded at an acceptable price.
Nuclear weapons permanently broke that calculation. They introduced a variable no military planner can rationally manage: the certainty of existential retaliation. A single surviving warhead, launched in response to an attack, can destroy a capital city within minutes. That possibility, not certainty but a credible possibility, fundamentally transforms how any potential aggressor models the cost of war.
This is the doctrine strategists call Mutually Assured Destruction, abbreviated as MAD. The logic is not complicated. If attacking a state guarantees the complete destruction of your own, the rational decision is to never attack. Both parties become deterred by the catastrophic consequences they would impose on each other. The hostility between them may remain. The invasion does not happen.
The product of this arrangement is not peace. It is an armed equilibrium, expensive, anxiety-inducing, and ethically troubling. Yet that equilibrium has held with remarkable consistency since 1945. No nuclear-armed state has directly invaded another nuclear-armed state in the 81 years since the first atomic bomb was used in warfare. That empirical record forms the foundation on which deterrence theory rests.
As the Nonproliferation Review's 2024 special issue on nuclear deterrence acknowledges, claims that deterrence averts conflict rest on reading non-events as success. Critics argue that framing downplays close calls and accidents, suggesting humanity has sometimes avoided nuclear catastrophe more by fortune than by design. That is a legitimate challenge, and this article addresses it directly.
You May Also Like
U.S.-Iran War: A 28-Day Full Breakdown
How deterrence failed, escalated, and reshaped global security in under a month.
Why This Matters: The Bigger Picture in 2026
Nuclear deterrence is not a Cold War artifact preserved in geopolitical textbooks. It is the active operating framework of contemporary international relations, shaping the decisions of every major power in every active conflict theatre right now.
Consider Ukraine. NATO collectively possesses substantial conventional military advantages over Russia across multiple capability categories. Yet the alliance has consistently avoided direct military engagement with Russian forces on Ukrainian territory. The reason is not indecision. It is nuclear geometry. As the Chatham House analysis of nuclear risk in 2026 documents, Russia has publicly and repeatedly linked the potential use of nuclear weapons to threats it perceives as existential. That signal changes NATO's risk calculus at every decision point. The conflict remains conventional in character precisely because nuclear weapons exist on one side of the equation and NATO's nuclear umbrella exists on the other.
Russia, simultaneously, avoids direct strikes against NATO member territory. The constraint operates in both directions. Both parties operate below a nuclear ceiling; neither is willing to test publicly. As reported in our coverage of the Islamabad peace talks collapsing with no U.S.-Iran deal, and the Iran ceasefire countdown reaching its April 21 threshold, the same deterrence dynamics are shaping multiple flashpoints simultaneously in 2026.
The Middle East, the Taiwan Strait, and the Korean Peninsula each follow the same structural pattern. Conventional pressure escalates through proxies, economic coercion, and political subversion. The nuclear ceiling prevents direct confrontation between major powers.
Global Nuclear Arsenals as of January 2026
All figures below come directly from the SIPRI Yearbook 2026. Precise warhead inventories remain classified by all nine nuclear-armed states. These are the most widely accepted open-source estimates in the international arms control research community.
As of January 2026, the total global inventory was 12,187 warheads, a marginal decline from 12,241 in January 2025. Of those, 9,745 were in military stockpiles considered potentially operationally available, up from 9,614 the previous year. An estimated 4,012 warheads were deployed with missiles and aircraft. Russia and the United States together account for approximately 86 percent of all warheads in existence.
| Country | Total Warheads (Jan 2026) | Strategic Role |
|---|---|---|
| Russia | 5,420 | Strategic parity and regional deterrence |
| United States | 5,042 | Global deterrence and extended alliance assurance |
| China | 620 | Expanding second-strike capability, the military stockpile grew from 600 to 620 in 2026 |
| France | 370 | Independent nuclear deterrent force; total reflects 80 retired warheads counted for the first time in 2026 |
| United Kingdom | 225 | Sea-based minimum deterrence via Vanguard-class submarines |
| India | 190 | Credible minimum deterrence doctrine |
| Pakistan | 170 | Regional deterrence balance against India |
| Israel | 90 (estimated) | Strategic ambiguity: Israel does not officially acknowledge its arsenal |
| North Korea | 60 (estimated) | Regime survival deterrence; assembled warhead count rose from 50 in 2025 |
Source: SIPRI Yearbook 2026, Chapter 6. Russia and the United States hold approximately 86 percent of all nuclear warheads in existence. Both continue extensive modernisation programmes. Global military stockpiles increased from 9,614 in January 2025 to 9,745 in January 2026, driven primarily by growth in Russia's deployed warhead count and China's expanding arsenal.
The Variable No One Covers: Second-Strike Survivability
Most media coverage of nuclear arsenals focuses almost entirely on warhead counts. That misses the variable that actually determines deterrence stability. The number that matters is not total warheads held. It is warheads capable of surviving a first strike and retaliating with certainty. This is second-strike capability, and it is the structural foundation on which MAD rests.
A state that can absorb a first strike and still retaliate devastatingly holds effective deterrence regardless of numerical gaps in total warhead counts. A state whose entire arsenal can theoretically be neutralised in a pre-emptive strike holds no deterrence at all, regardless of how many warheads it possesses on paper.
This is why submarine-launched ballistic missiles are the most strategically valuable component of any nuclear arsenal. A nuclear-armed submarine at an undisclosed depth in an ocean of unknowable location is, for practical purposes, unstrikeable. The United States cannot guarantee the destruction of all Russian submarines in a preemptive strike. Russia cannot guarantee the same against the United States. Both sides know retaliation is certain, and that certainty is the mechanism that prevents the first strike from ever being ordered.
As the Brookings Institution's post-New START analysis notes, command stability, communication survivability, and second-strike credibility collectively matter far more to deterrence effectiveness than raw warhead numbers. China's 600 warheads are deliberately diversified across land-based missiles, submarine-launched systems, and aircraft. The objective is not numerical parity. It is guaranteed second-strike capability, and therefore guaranteed deterrence regardless of what an adversary attempts first.
The SIPRI Yearbook 2025 flags that quantum detection technology and advances in anti-submarine warfare could theoretically erode submarine survivability. If that capability matures to operational deployment, it would destabilise the very foundation of MAD. That is a long-term risk, not a present reality, but it is the most serious technological challenge deterrence researchers are currently tracking.
You May Also Like
AI in Warfare: What It Can Actually Do in 2026
Autonomous targeting, real-time surveillance, and the weapons systems are already operational right now.
Eight Decades of Evidence: The Historical Record Since 1945
The Cold War provides the clearest empirical test of deterrence theory available in modern history. The United States and the Soviet Union competed militarily across every continent, funded opposing sides in proxy wars from Korea to Angola to Afghanistan, and came to within minutes of nuclear exchange during the Cuban Missile Crisis of October 1962. Neither side fired a single nuclear weapon at the other across the entire 45-year duration of that confrontation.
The restraint was not the product of mutual affection or shared values. Both powers deeply distrusted each other and spent enormous resources planning for nuclear war. The restraint was the product of calculation: a direct attack guarantees a response that destroys you. That calculation held under extraordinary pressure, through multiple crises, and across multiple administrations on both sides. The RAND Corporation's extended deterrence research documents this pattern as consistent across multiple regional security dyads throughout the Cold War and post-Cold War periods.
India and Pakistan fought three conventional wars before either state acquired nuclear weapons. Following their reciprocal nuclear tests in May 1998, the character of their conflicts changed. The 1999 Kargil conflict saw Pakistani-supported forces occupy Indian positions in the disputed territory at an altitude. India possessed the conventional military capacity to expand operations significantly across the international border. It did not do so. Pakistan's nuclear status imposed a ceiling on India's response options that no combination of diplomatic pressure or conventional military capability could replicate alone. Both sides de-escalated.
More recently, SIPRI researcher Matt Korda warned in June 2025 that the brief conventional conflict between India and Pakistan in early 2025 risked turning into a nuclear crisis when strikes hit nuclear-related military infrastructure, describing it as a stark warning for states seeking to increase their reliance on nuclear weapons.
The CFR Global Conflict Tracker confirms that every active conflict today involving nuclear-armed states remains either proxy-based or deliberately constrained in scope.
Nuclear Deterrence vs Conventional Deterrence
A recurring argument holds that conventional military strength, sanctions, and diplomacy can replicate the deterrence effect of nuclear weapons. The historical record does not support that claim. As the MIT Press journal International Security concluded in 2025, conventional deterrence is generally weaker than nuclear deterrence, though emerging conventional capabilities are beginning to assume a larger role in deterring limited nuclear use in specific scenarios.
| Aspect | Nuclear Deterrence | Conventional Deterrence |
|---|---|---|
| Effectiveness against major invasion | Prevents direct attack between nuclear states via existential threat | Limited by force ratios; vulnerable to miscalculation under pressure |
| Escalation risk | Catastrophic if used; restraint has historically prevailed between nuclear states. | Moderate to high; may provoke proportional or escalatory responses |
| Cost structure | High initial investment; comparatively low recurring cost once the arsenal is credible | Continuous high expenditure on personnel, logistics, and equipment |
| Psychological impact | Deters adversaries through guaranteed destruction logic and second-strike uncertainty | Perceived strength may not match actual battlefield capability |
| Strategic flexibility | Limited by design; primarily serves deterrence rather than warfighting | High; applicable across the full spectrum from skirmish to war |
| Treaty verification | Historically governed by bilateral inspection and data-exchange regimes | Generally unregulated and unverifiable by neutral third parties |
Legitimate Criticisms: Where the Deterrence Argument Gets Complicated
The case for nuclear deterrence as a stabilising force carries genuine and unresolved weaknesses that any credible analysis must address directly.
The most fundamental challenge is the irrational actor problem. Deterrence rests entirely on the assumption that all parties value regime survival and behave accordingly. North Korea represents the sharpest available test of that assumption. SIPRI estimates that North Korea has assembled approximately 50 warheads. The deterrence calculus is openly transactional: Saddam Hussein did not possess nuclear weapons and was removed by external force. Muammar Gaddafi surrendered his nuclear programme in 2003 and was killed in a NATO-backed uprising eight years later. Kim Jong Un observed both outcomes and drew the obvious conclusion. North Korea's arsenal exists to make the cost of regime change so high that no external power will attempt it. That arrangement has held for over two decades.
Iran has studied the same pattern. The ongoing question of Iranian nuclear capability sits at the centre of regional deterrence dynamics throughout the Middle East. As detailed in our reporting on how the U.S.-Iran confrontation unfolded across 28 days, a non-nuclear state navigating a world structured around nuclear deterrence faces fundamentally different strategic pressures than one that has crossed the threshold.
The near-miss problem is equally serious. According to documented research cited by ICAN, multiple incidents since 1962 involved technical failures, communication breakdowns, and human error that came close to triggering unauthorised or accidental launches. The most documented: Soviet officer Stanislav Petrov's 1983 decision to disregard a false alarm from a faulty satellite system that reported an incoming U.S. first strike. The SIPRI Yearbook 2025 acknowledges that the system has survived these incidents while warning that structural stability cannot be assumed to be indefinitely self-sustaining.
The proliferation risk is the third major vulnerability. Every additional state that acquires nuclear weapons introduces new command-and-control variables, new potential irrational actors, and new pathways through which weapons or fissile material could reach non-state actors.
You May Also Like
Small, Cheap, and Deadly: The New Face of Modern Warfare
Why drones, precision munitions, and asymmetric weapons are rewriting every military doctrine on earth.
The Command Chain Nobody Talks About: Who Actually Has the Authority
Almost every article on nuclear deterrence stops at warhead counts and MAD theory. The actual decision architecture, who holds physical authority, what protocols govern a launch, and where those protocols can break down, gets almost no coverage in general-audience reporting. That gap matters because the stability of deterrence depends not just on whether weapons exist, but on whether the humans and systems controlling them will behave predictably under extreme pressure.
The "nuclear football" is not a launch button.
The U.S. Presidential Emergency Satchel, widely called the "nuclear football," contains authentication codes, not a trigger. A launch order flows from the President through the National Military Command Center, then to Strategic Command, and finally to submarine crews and missile launch officers through multiple authentication steps. Two separate people at every terminal must independently confirm the order. The failure points are in the authentication chain, the communication links, and the time pressure applied to the humans at the end of it.
Russia's semi-automated retaliation system
Russia operates a semi-automated retaliatory system, officially designated Perimeter, known in the West as "Dead Hand." If sensors detect a nuclear strike and communication with the Russian high command is disrupted, the system is designed to transmit launch commands without further human authorisation. Its current operational status is unconfirmed by Russian officials, but Western defence analysts widely assume it remains active. Its existence means Russian second-strike capability does not depend entirely on the survival of rational decision-makers. Deterrence theory assumes human actors at the end of the chain. Perimeter removes that assumption for at least one side of the largest nuclear dyad in the world.
Pakistan's command authority is genuinely disputed.
Pakistan's Strategic Plans Division controls its nuclear arsenal, but the civilian government's actual authority over a launch decision in a crisis has never been publicly verified and is disputed in the scholarly literature. In a conflict scenario involving rapid escalation with India, the identity of the decision-maker and the timeline of that decision are not publicly known quantities. Arms control researchers flag this as one of the most serious gaps in current deterrence analysis.
The sole authority problem in the United States
Under current U.S. law, a president can order a nuclear strike unilaterally. No Congressional authorisation is required. No Secretary of Defense approval is legally mandated. Legislation to require a second authorising official has been proposed multiple times and has not passed. Former nuclear launch officers and military lawyers have flagged this as a structural vulnerability. The system was designed under the assumption that speed of response mattered more than deliberation.
The Stability-Instability Paradox: How Nuclear Weapons Enable Some Wars
The most important strategic concept almost entirely absent from general-audience nuclear coverage is the stability-instability paradox. It explains why the 81-year record of no direct nuclear-state-to-nuclear-state war, accurate as a statistic, tells only part of the story.
The paradox works like this. Nuclear deterrence stabilises the highest level of conflict, direct major-power war, while simultaneously making lower-level aggression cheaper and safer for the party using it. If your nuclear arsenal prevents your adversary from escalating past a certain threshold, you can conduct limited military operations below that threshold with reduced risk of a catastrophic response. The nuclear ceiling does not just cap escalation. It creates a protected space underneath it.
Kargil 1999: the clearest documented case
Pakistan's acquisition of nuclear weapons made the Kargil incursion of 1999 strategically viable for Pakistani military planners. They calculated that India's conventional military advantage was neutralised at the escalation ceiling. India could not cross the international border without risking nuclear escalation. Pakistani forces held their positions for months before international pressure forced their withdrawal. Nuclear deterrence prevented a full-scale India-Pakistan war. It also made Kargil possible. Those two facts are the same mechanism producing different outcomes at different levels of conflict.
Russia in Ukraine: the paradox in real time
Russia's nuclear posture did not prevent its invasion of Ukraine. It constrained NATO's response options at every stage. Crimea in 2014, the Donbas proxy conflict, and the full-scale 2022 invasion all took place under an explicit nuclear umbrella. Each time NATO considered a response, Russia's nuclear signals set the ceiling on what was possible. Russian conventional aggression below the nuclear threshold proceeded at a pace and scale that would have been far less likely if both sides were non-nuclear.
China and Taiwan: grey-zone operations under the umbrella
China's air incursions into Taiwan's Air Defence Identification Zone have intensified in frequency alongside its nuclear arsenal expansion. The nuclear ceiling does not prevent those incursions. It makes the cost of a direct military response by the United States more complex to calculate. Sub-threshold coercion becomes cheaper when the upper end of the escalation ladder is blocked.
Image Credit: Leonardo AI
Myth vs Reality: What Deterrence Theory Actually Claims
Nuclear deterrence is one of the most widely misunderstood strategic concepts in public discourse. The myths exist on both sides of the argument. Its defenders and its critics both regularly argue against a version of the theory that no serious strategist actually holds.
| The Myth | The Reality | Source |
|---|---|---|
| Deterrence requires numerical parity | It requires second-strike survivability, not equal warhead counts. France has 290 warheads. The UK deters with 225, all submarine-based. Past the credible retaliation threshold, additional warheads add marginal deterrence value and substantial political cost. | FAS / SIPRI 2025 |
| More warheads mean more deterrence. | The U.S. held over 30,000 warheads at its Cold War peak. Its current 5,177 provides effectively identical deterrence credibility. Past a minimum threshold, more warheads produce no additional deterrence effect. | Federation of American Scientists, 2025 |
| Russia's invasion of Ukraine proves deterrence failed. | Deterrence theory never claimed to prevent a nuclear-armed state from invading a non-nuclear state. Ukraine surrendered its Soviet-era weapons under the Budapest Memorandum in 1994 in exchange for security assurances, not a mutual deterrence relationship. Russia's invasion is exactly the outcome deterrence theory predicts when one party removes the nuclear variable. | Arms Control Association, 2025 |
| Nuclear weapons prevent all wars. | They prevent a specific kind of war: direct large-scale interstate conflict between nuclear-armed states, and only when second-strike capability is credible. They have no demonstrated effect on civil wars, terrorism, insurgencies, or proxy conflicts. | RAND Corporation, 2023 |
| The Cuban Missile Crisis proves that deterrence works cleanly. | It does, but the resolution depended on a secret diplomatic back-channel (the U.S. quietly removing Jupiter missiles from Turkey), Soviet submarine officer Vasili Arkhipov refusing to authorise a nuclear torpedo without Moscow's confirmation, and the U.S. not retaliating when a reconnaissance aircraft was shot down. Three individuals making three separate judgment calls prevented a nuclear exchange. That is deterrence surviving on narrow margins, not deterrence working by design. | National Security Archive, declassified 2002 |
The Extended Deterrence Problem: When Allies Start Doubting the Umbrella
Extended deterrence is the arrangement by which a nuclear power extends its nuclear guarantee to non-nuclear allies. The United States does this formally for NATO members, Japan, South Korea, and Australia. The structural problem with every such arrangement is the same: credibility. Would the United States actually trade New York for Seoul? Every American ally privately asks that question. The answer is not self-evident, and when it starts to look uncertain, the consequences for proliferation and regional stability can be severe.
South Korea: the clearest credibility gap in the system
Multiple polls conducted between 2022 and 2025 found that between 60 and 70 percent of South Koreans support developing an independent nuclear capability. That number does not reflect aggression. It reflects doubt about whether the U.S. extended deterrence guarantee holds in a scenario where North Korea's intercontinental ballistic missiles make a Korean peninsula conflict directly costly to the American homeland. South Koreans have noticed the ambiguity and are responding accordingly in polling data.
Japan's structural bind
Japan is in a structurally identical position with a compounding constraint. Its constitution's war-renouncing clause and the moral weight of Hiroshima and Nagasaki make any domestic nuclear debate politically difficult in a way that applies to no other U.S. ally. Japan relies entirely on the U.S. umbrella with no independent option and a constrained domestic environment for building one. If the umbrella weakens, Japan's choices narrow considerably.
NATO nuclear sharing: the physical credibility mechanism
NATO's nuclear sharing arrangement places U.S. B61 gravity bombs in Belgium, Germany, Italy, the Netherlands, and Turkey. The weapons are there partly as a tangible signal that the U.S. commitment is not just diplomatic language. The arrangement is legally complex: host nations' aircraft would deliver U.S. weapons under NATO authorisation, but nuclear release authority remains American. Germany's recurring internal debate about whether to continue participating is worth tracking as a credibility indicator. Each time it resurfaces, it signals that allies are stress-testing the guarantee rather than taking it for granted.
Political unpredictability as a deterrence variable
The 2024-2025 period saw multiple U.S. allies quietly accelerate extended deterrence consultations with Washington. The concern was not that the United States would deliberately abandon allies. It was that political unpredictability at the presidential level creates enough ambiguity to erode deterrence credibility at the margins. In deterrence, margins are where the system either holds or fails. An ally that is 90 percent confident in the guarantee behaves differently from one that is 60 percent confident, and that difference shows up in defence spending, diplomatic positioning, and proliferation pressure.
Escalation Dominance vs Minimum Deterrence: The Doctrine Debate Driving the Next Arms Race
This section addresses the strategic doctrine debate actively happening inside defence establishments right now. It rarely surfaces in public reporting because it requires familiarity with actual nuclear war-planning logic rather than just strategic theory. The outcome of this debate will determine whether the post-New START world stabilises at a new equilibrium or enters a genuine three-way arms race.
The two doctrines are stated plainly.
Minimum deterrence holds that a small, survivable arsenal with guaranteed second-strike capability is sufficient and strategically optimal. France and the United Kingdom operate on this doctrine. The argument is that once you can guarantee retaliation, additional warheads add no deterrence value and create substantial political and verification costs.
Escalation dominance holds that deterrence requires the ability to respond credibly at every level of the escalation ladder. Under this doctrine, a state needs tactical nuclear options for battlefield use, theatre nuclear options for regional conflicts, and strategic options for existential threats, so that an adversary cannot threaten limited nuclear use without facing a credible limited nuclear response at the same level.
Why the distinction matters in 2026
The U.S. B61-12 gravity bomb, with its new precision tail kit enabling lower-yield accurate delivery, reflects escalation dominance logic. Minimum deterrence doctrine considers weapons like this actively destabilising, because they make nuclear use more thinkable at the lower rungs of the escalation ladder, which removes the firebreak between conventional and nuclear conflict.
Russia's official nuclear doctrine, updated in November 2024, explicitly expands the range of contingencies under which Russia could use nuclear weapons. This reflects escalation dominance logic that the U.S. and NATO minimum deterrence posture was not designed to counter at the lower rungs. (SIPRI Yearbook 2025, Chapter 6)
China's stated doctrine vs its revealed behaviour
China maintains an official "no first use" doctrine, the clearest stated commitment to minimum deterrence among major nuclear powers. Its expansion from 200 to 600 warheads does not necessarily contradict that doctrine if the growth is about guaranteeing second-strike survivability at scale. It starts to contradict it if China builds toward 1,500 warheads and theatre nuclear options by 2035, which would represent a quiet shift toward escalation dominance without declaring one. The U.S. Department of Defense's trajectory projections suggest that a shift may be underway.
The trilateral arithmetic that drives the next arms race
U.S. nuclear force posture was sized around bilateral deterrence with Russia. If China reaches 1,000 warheads by 2030 and continues building, U.S. planners face pressure to size forces against two near-peer nuclear adversaries simultaneously. At that point, minimum deterrence arithmetic stops producing confident answers. Escalation dominance logic then demands a substantially larger arsenal. That is the mechanism through which a three-way arms race begins: not through deliberate aggression, but through planning mathematics applied under conditions of strategic uncertainty. As Brookings noted in February 2026, China's expansion is the dominant factor that no existing arms control framework was designed to address.
| Dimension | Minimum Deterrence | Escalation Dominance |
|---|---|---|
| Arsenal-sized target | Smallest survivable second-strike force | Full-spectrum capability across all conflict levels |
| Tactical nuclear options | Unnecessary; lowers the threshold for use | Required; closes gaps at lower escalation rungs |
| First-use posture | No first use or ambiguous; retaliatory focus | First use permissible at lower thresholds as a de-escalation tool |
| Arms race risk | Low; capped by sufficiency logic | High; adversary parity logic drives continuous expansion |
| Treaty compatibility | High; small arsenals are easier to verify and limit | Low; tactical weapons are harder to count and verify |
| Who uses it now | France, UK, China (stated) | Russia (updated 2024 doctrine), U.S. (partially, via B61-12) |
What Happens Next: Three Verified Stress Points in 2026
The deterrence architecture has functioned for eight decades. Three concurrent developments in 2026 are placing simultaneous structural stress on its foundations in ways the Cold War framework was not designed to accommodate.
China's Arsenal Expansion Creates an Unprecedented Trilateral Problem
China's nuclear arsenal stood at approximately 200 warheads in 2019. By January 2025, SIPRI confirmed it had reached 600, growing at approximately 100 new warheads per year since 2023. By January 2025, China had completed or was close to completing around 350 new ICBM silos. The U.S. Department of Defense projects that China could hold at least 1,000 operational nuclear warheads by 2030, and potentially 1,500 by 2035. (SIPRI, June 2025)
The United States now confronts a strategic environment it has never previously navigated: genuine trilateral nuclear competition among three states with comparable second-strike capabilities, with no established diplomatic framework for managing it. All existing deterrence doctrine, arms control treaties, and verification mechanisms were designed for the bilateral U.S.-Soviet relationship.
New START Expired on 5 February 2026 with No Replacement
The New Strategic Arms Reduction Treaty, which limited both the United States and Russia to 1,550 deployed strategic nuclear warheads and established an extensive verification regime including on-site inspections and biannual data exchanges, officially expired on 5 February 2026. As the Council on Foreign Relations confirmed, this is the first time since the early 1970s that the world's two largest nuclear arsenals operate without any legally binding numerical limits, mandatory inspection rights, or formal transparency mechanisms.
Russia proposed a one-year voluntary extension of numerical limits. President Trump responded on social media on 5 February that "we should have our Nuclear Experts work on a new, improved and modernized Treaty," but did not commit formally before the deadline. UN Secretary-General António Guterres urged both sides to immediately conclude a new agreement. No successor talks had begun as of the date of this article. (Nuclear Threat Initiative, January 2026)
As FAS explained in February 2026, when transparency collapses, both sides must rely on intelligence estimates and worst-case planning. In a crisis, that creates pressure to launch on warning, forcing decisions within minutes rather than hours, before an incoming attack is fully confirmed. That is the condition the arms control architecture of the past five decades was designed to prevent. That architecture no longer exists.
The strategic reality is documented in detail in our analysis of nuclear deterrence in 2026 with no treaty and no rules.
Artificial Intelligence Compresses Decision Time Below Safe Thresholds
The same artificial intelligence systems transforming conventional battlefield operations are being integrated into nuclear command, control, and communications infrastructure by multiple nuclear-armed states simultaneously. As covered in our reporting on what AI can actually do in modern warfare, these are not theoretical future scenarios. They describe operational systems already deployed or in advanced development.
The stability of Mutually Assured Destruction historically depended on sufficient decision time for political leaders to verify an attack signal, assess its credibility, consult advisors, consider alternatives, and choose a measured response. The Cuban Missile Crisis of 1962 took 13 days to resolve. AI integration into early warning systems reduces the decision window to minutes or seconds. The SIPRI Yearbook 2025 specifically flags AI and autonomous systems as a potential trigger for nuclear war through automated misidentification of a threat signal.
There are no established international norms, treaties, or confidence-building measures governing AI integration into nuclear systems in any of the nine nuclear-armed states as of May 2026.
Key Verified Facts at a Glance
- 12,241 nuclear warheads globally as of January 2025 (SIPRI Yearbook 2025, confirmed)
- 9,614 warheads in military stockpiles are considered potentially operationally available
- Approximately 2,100 warheads are at high operational alert on ballistic missiles, nearly all belonging to Russia or the United States
- Zero nuclear weapons used in warfare since Nagasaki, Japan, 9 August 1945
- Russia (5,459) and the United States (5,177) together hold approximately 90 percent of all warheads
- China's arsenal grew by approximately 100 warheads per year since 2023, from 200 in 2019 to 600 in January 2025
- New START, the last nuclear arms control treaty between the U.S. and Russia, expired on 5 February 2026 with no successor agreement
- For the first time since the early 1970s, the U.S. and Russian arsenals operate with no legally binding numerical limits
- No direct military conflict has occurred between two nuclear-armed states in 81 years
- In early 2025, India and Pakistan's conventional conflict briefly struck nuclear-related military infrastructure, per SIPRI June 2025
- Second-strike survivability, not aggregate warhead totals, is the variable that determines whether deterrence remains credible
- No international norms govern AI integration into nuclear early warning or targeting systems in any of the nine nuclear-armed states
An Uncomfortable Safeguard: What the Evidence Actually Tells Us
Nuclear weapons are simultaneously humanity's most destructive invention and, by the weight of eight decades of verified evidence, one of the most effective constraints ever placed on direct large-scale war between major powers. That paradox does not resolve neatly. It sits, unresolved and profoundly uncomfortable, at the centre of the international security order.
The case for nuclear deterrence as a stabilising force is not that it produces peace. It clearly does not. Wars happen constantly beneath the nuclear ceiling: proxy conflicts, regional invasions, civil wars, insurgencies, and state-sponsored terrorism. The case is narrower and more specific. No nuclear-armed state has directly invaded another nuclear-armed state in 81 years. That singular empirical fact, consistent across every geopolitical configuration since 1945, constitutes substantial evidence that the deterrence mechanism functions broadly as theorised.
The question relevant to 2026 is not whether deterrence has worked historically. The evidence is reasonably clear that it has. The relevant question is whether an architecture built on Cold War bilateral logic, mutual verification treaties, and rational actor assumptions can hold in a world characterised by trilateral nuclear competition, no binding arms control agreements, accelerating proliferation pressure, and AI integration into nuclear command systems. As the American Friends Service Committee concluded in February 2026, when transparency collapses, nuclear-armed states opt for worst-case thinking, and that increases pressure to build more weapons and signal more aggressively. At its best, arms control lowers the risk that misunderstanding becomes escalation. That best-case scenario no longer has a legal architecture supporting it.
The rules that governed the first eight decades of the nuclear age kept major-power direct warfare from happening. The rules that will govern the next eight decades have not yet been written, and the window in which to write them is narrowing faster than most governments publicly acknowledge.
The System Has Worked. The System Is Also Fragile.
The facts, read plainly, support one conclusion: nuclear deterrence has prevented direct major-power war for 81 years. That is not an argument for nuclear weapons. It is an observation about what has happened. No serious analyst disputes the 81-year record. The disagreement is about what comes next and whether the mechanisms that produced that record are still intact.
The honest answer is that some of them are not. The bilateral U.S.-Russia arms control architecture is gone. The framework that kept both sides from worst-case planning no longer has a legal structure behind it. China's build-up creates a three-way strategic competition that no existing doctrine or treaty was designed for. Russia's updated nuclear doctrine in November 2024 expanded the range of contingencies for nuclear use. AI is compressing the time available for human judgment in exactly the scenarios where human judgment has historically mattered most. These are documented changes to the operating environment, not hypothetical risks.
The deterrence skeptics are right that the system has survived on narrow margins before. Vasili Arkhipov in a Soviet submarine in 1962. Stanislav Petrov at a Soviet early warning post in 1983. Each time, one person made one call that the system's design did not guarantee they would make correctly. Deterrence theorists are also right that the overall record is more consistent than critics acknowledge, and that removing nuclear weapons without replacing them with a credible alternative would not automatically produce a safer world. It might produce a world where major-power conventional war becomes possible again for the first time since 1944.
What the evidence does not support, from either side of the debate, is confidence. The system has held. The system is under more simultaneous structural stress right now than at any point since the early 1980s. Both of those things are true. Anyone telling you only one of them is giving you a partial picture, and partial pictures are how serious miscalculations begin.
If nuclear weapons disappeared overnight, every warhead and every delivery system gone, would the world become substantially safer? Or would the absence of that ultimate deterrent simply remove the ceiling on conventional warfare between major powers, making large-scale war between great powers possible again for the first time since 1945? That question has no comfortable answer. Share this article with someone who disagrees with your instinct and find out where the argument leads.