Nuclear risk: where do we stand?

Nuclear risk is the reason the Doomsday Clock exists, and it carries the heaviest weights in the simulator: it is the only scenario that can flip the world in under an hour. After thirty years of decline, it has been rising since 2019. Here is where things stand, without sensationalism.

How many weapons, and who has them?

Roughly 12,100 nuclear warheads exist in early 2026 according to the Federation of American Scientists, of which ~9,600 are in military stockpiles and ~3,900 deployed. Russia (~5,500) and the United States (~5,000) hold nearly 90%. Then come China (~600, growing fast), France (290), the United Kingdom (225), India, Pakistan, Israel and North Korea. At the Cold War peak (1986) there were 70,000: the decline was real, but it has stopped, and China is the first state to significantly expand its arsenal in thirty years.

What is a nuclear winter?

Explosions over cities would ignite firestorms whose soot would rise into the stratosphere, above the rain that could wash it out. It would stay there for years, blocking sunlight. A landmark study (Xia et al., Nature Food, 2022) estimates that a US–Russia war involving 4,400 warheads would cut temperatures by several degrees and global crop production by 90% in year two: over 5 billion deaths from famine. Even a 'limited' India–Pakistan war (100 warheads) would kill up to 2 billion people through its climate effects. Nuclear winter, not the blasts, is what makes nuclear war a civilisational risk.

Why is the risk rising?

• The end of treaties. ABM (2002), INF (2019), Open Skies (2020), and New START expired in February 2026 with no successor: for the first time since 1972 no agreement caps US and Russian arsenals.
• Explicit threats. Since 2022 Russia has repeatedly raised the use of nuclear weapons over Ukraine and revised its doctrine to lower the threshold.
• Modernisation. All nine nuclear states are renewing their arsenals: ~$100 billion a year (ICAN, 2024).
• New technologies. Hypersonic missiles, AI in early-warning systems, cyberattacks on command and control: all shrink decision time — and raise the risk of error.

The near misses

The risk is not only deliberate war, it is error. In 1983 Soviet officer Stanislav Petrov judged, alone, that a US launch warning was a false alarm (it was sunlight glinting off clouds). In 1995 a Norwegian scientific rocket was mistaken for a Trident missile and the Russian nuclear briefcase was activated. In 1979 and 1980, a training tape and a faulty chip triggered false alerts in the United States. Every case was resolved by a human who doubted.

What has already turned the clock back

Nine moves back since 1947, all tied to arms control: the Partial Test Ban (1963), NPT (1969), SALT I (1972), INF (1988), START I (1991), New START (2010). The lesson is plain: treaties work. They took arsenals from 70,000 to 12,000 warheads. A successor to New START, a finally ratified Comprehensive Test Ban Treaty, or new states joining the Treaty on the Prohibition of Nuclear Weapons (TPNW, 2021, 73 states parties) would be the strongest signals.

And at my scale?

Nuclear weapons are the area where individual action feels most futile. Three things nonetheless count: where your bank and pension are invested (the Don't Bank on the Bomb campaign lists nuclear weapons financiers); your representatives' stance on the TPNW and the deterrence budget; and your support for the organisations doing the expert work (ICAN, 2017 Nobel Peace Prize; the Bulletin; the Arms Control Association). See Act.