No one can predict whether nuclear war will happen, but the risk is real and, by several expert measures, higher now than at any point since the Cold War. As of January 2026, the Bulletin of the Atomic Scientists set its Doomsday Clock to 85 seconds to midnight, the closest it has ever been. That assessment reflects a combination of expanding arsenals, collapsing arms control agreements, active regional conflicts between nuclear-armed states, and new technological risks that didn’t exist a decade ago.
Understanding whether nuclear war is likely means looking at what would have to go wrong, how close the world has come before, and what safeguards remain. The short answer: a full-scale nuclear exchange is not inevitable, but the barriers preventing one are thinner than most people realize.
The Current State of Nuclear Weapons
Nine countries possess nuclear weapons, with a combined global inventory of roughly 12,241 warheads as of January 2025. Of those, about 9,614 are in active military stockpiles, meaning they could be prepared for use. The rest are retired warheads awaiting dismantlement. Russia holds the largest arsenal at about 5,459 total warheads, followed by the United States at 5,177. Together, they account for nearly 87% of all nuclear weapons on Earth.
The overall number of warheads has been declining for years, but that’s entirely because the U.S. and Russia are slowly dismantling old, retired weapons. Meanwhile, several other countries are actively building up. China’s stockpile grew from an estimated 500 warheads in January 2024 to 600 in January 2025, and it’s expected to keep expanding over the next decade. India, Pakistan, and North Korea are all increasing their arsenals as well. The United Kingdom announced plans to raise its warhead cap from 225 to 260. So while the total count ticks down, the number of countries with growing nuclear capability is going up.
Why the Risk Is Elevated Right Now
Several factors are converging in ways that concern security analysts. The New START treaty, the last major arms control agreement between the U.S. and Russia limiting deployed nuclear weapons, was extended only through February 4, 2026. Russia suspended its participation in the treaty in 2023, and no successor agreement is in place. Without a framework for mutual inspections and warhead limits, both countries lose visibility into what the other is doing, which breeds suspicion and incentivizes building more weapons rather than fewer.
Regional flashpoints add another layer of danger. The 2025 crisis between India and Pakistan demonstrated how quickly two nuclear-armed neighbors can approach the edge. Analysts at Harvard’s Belfer Center noted that Pakistan invoked its right to self-defense under international law while India struck sensitive military sites, pushing the region “perilously close to the edge.” The geography between these two countries is compressed, military alert postures are high, and there is virtually no buffer zone where a conflict could be fought and contained once certain thresholds are crossed. The idea of a “limited nuclear war” between India and Pakistan, experts concluded, is a dangerous illusion.
Russia’s war in Ukraine has also reintroduced nuclear threats into mainstream geopolitics in a way not seen in decades. Russian officials have repeatedly referenced their nuclear capability in the context of the conflict, lowering the rhetorical bar for when nuclear use might be considered.
How Nuclear War Could Start by Accident
One of the most underappreciated risks is that nuclear war could begin not from a deliberate decision, but from a mistake. History is full of near misses. During the Cuban Missile Crisis in 1962, a guard at a military base in Minnesota saw someone climbing a security fence and triggered a sabotage alarm. The alarm cascaded to a nearby base in Wisconsin, where a faulty system sent nuclear-armed fighter jets scrambling into the air. The intruder turned out to be a bear. During the Suez Crisis, NORAD received simultaneous reports of unidentified aircraft over Turkey, Soviet fighters over Syria, a downed British bomber, and unexpected Soviet naval movements, all of which together looked like the opening moves of a Soviet attack but weren’t.
These incidents happened when communication was slower and weapons systems were simpler. Today, the window for making decisions is even narrower. A nuclear-armed intercontinental ballistic missile can travel from Russia to the United States in about 30 minutes, and submarine-launched missiles can arrive in as little as 10 to 15 minutes. That leaves very little time to confirm whether a warning is real before a leader must decide whether to launch a retaliatory strike.
Artificial Intelligence as a New Risk Factor
The integration of AI into nuclear command systems introduces risks that didn’t exist during the Cold War. The U.S. nuclear command and control architecture is a complex system of over 200 components, and simpler forms of AI are already being used in early warning sensors. As AI takes on more responsibility in detecting threats and supporting decisions, several problems emerge.
Generative AI systems can “confabulate,” confidently presenting false information as though it were real. In a low-stakes environment, that’s a nuisance. In a system designed to detect incoming nuclear missiles, it could be catastrophic. There’s also the problem of automation bias: people tend to trust AI-generated analysis more than they should, especially when the system mimics human-like reasoning. Over time, this can shift a human decision-maker from actively choosing whether to act to passively watching an AI make recommendations, a subtle but dangerous erosion of human control.
On a strategic level, AI-powered surveillance could allow one country to track another’s hidden nuclear submarines or mobile missile launchers, potentially undermining the other side’s ability to retaliate after a first strike. If a country believes its second-strike capability is compromised, it has a strong incentive to launch first in a crisis rather than risk losing its arsenal. That dynamic could fuel arms races and make conflicts harder to de-escalate.
What a Nuclear War Would Actually Mean
The consequences of even a “limited” nuclear exchange would be staggering. U.S. government modeling of attacks targeting only industrial or military sites (not cities) with a combined yield of 100 megatons estimated more than 10 million deaths. A larger counterforce attack targeting U.S. nuclear forces, involving roughly 3,000 warheads, could kill between 13 and 34 million people and cause 25 to 64 million total casualties, including severe injuries and radiation illness. When fire and firestorm effects are factored in, some models predict death tolls 1.5 to 4 times higher than blast-only estimates, potentially reaching 56 million.
Survivors in fallout zones would face acute radiation sickness. Without intensive medical treatment, a radiation dose of around 250 to 450 rads is enough to kill half of those exposed within 60 days. In a large-scale exchange, medical infrastructure would be destroyed in exactly the areas where it was most needed. These numbers don’t account for longer-term effects like nuclear winter, crop failure, or the collapse of supply chains, which could threaten billions of people globally even in regions far from any detonation.
What Keeps Nuclear War From Happening
The primary mechanism preventing nuclear war is deterrence: the understanding that any country launching a nuclear attack would face devastating retaliation. This logic has held since 1945, and it remains the foundation of nuclear strategy for every armed state. No rational leader would order a first strike knowing their own country would be destroyed in response.
But deterrence has limits. It assumes rational decision-making, clear communication, accurate information, and functioning technology. Every near miss in history represents a moment when one or more of those assumptions almost failed. Deterrence also doesn’t protect against non-state actors, though no terrorist organization is currently known to possess a nuclear weapon.
Diplomatic channels, hotlines between nuclear-armed states, and international institutions like the International Atomic Energy Agency all serve as additional buffers. Arms control treaties, when they exist, provide transparency and predictability. The erosion of those treaties is precisely why experts are worried: not because leaders want nuclear war, but because the systems designed to prevent accidents and miscalculations are weaker than they’ve been in decades.
How Likely Is It, Really?
No serious analyst assigns a precise probability to nuclear war, but the expert consensus is that the risk is low in any given year and unacceptably high over time. A 1% annual probability of nuclear war, a figure some researchers have used as a rough estimate, translates to roughly a 26% chance over 30 years. The actual probability is impossible to calculate because it depends on human decisions, political crises, and technical failures that can’t be modeled with precision.
What’s clear is that the trend lines are moving in the wrong direction. More countries are building more weapons. The last major arms control treaty is expiring without a replacement. AI is being integrated into warning systems faster than safety frameworks can keep up. And active conflicts between nuclear-armed states are testing deterrence in real time. Nuclear war is not likely on any given day, but the conditions that make it possible are more present now than they have been in a generation.