The independent resource on global security

Estimating world nuclear forces: An overview and assessment of sources

Estimating world nuclear forces
Sabre sunrise. Photo: US Dept of Defense

The nuclear weapon inventories of the nine-nuclear armed states—the United States, Russia, the United Kingdom, France, China, Israel and the Democratic People’s Republic of Korea (DPRK or North Korea)—are largely shrouded in secrecy: only three of the nine states have ever publicly declared the size of their nuclear stockpiles and in recent years there has been a notable shift towards a lower overall level of nuclear weapon-related transparency.[1] This means that any assessments of the size and composition of global nuclear arsenals are estimates. Such estimates are fraught with uncertainty because of the scarcity of authoritative sources and because of the nature of the methods used to analyse the limited data that is publicly available. This Topical Backgrounder outlines and evaluates the sources used to produce SIPRI’s annual estimates of world nuclear forces published in the SIPRI Yearbook.

Lack of transparency as a key challenge

Some nuclear-armed states publish weapon programme budgets and descriptions; however, governments rarely disclose comprehensive information about their nuclear arsenals. Only the USA, the UK and France have previously made statements about the size of their nuclear arsenals. Of these three, the USA has been the most transparent with the declassification of nearly the entire history of the size of its nuclear weapon stockpile. France and the UK have occasionally declared the approximate numbers of the nuclear weapons they possess. The other six nuclear-armed states have never publicly disclosed the size of their nuclear arsenals, and one of these states (Israel) does not even publicly acknowledge its own possession of nuclear weapons.

Unfortunately, in recent years both the USA and the UK have begun to revert to cold war-era levels of nuclear opacity. In the USA, for example, the administration of President Donald J. Trump reversed its predecessor’s transparency policies and refused to declassify the size of the US nuclear military stockpile and the number of dismantled warheads in both 2019 and 2020. Similarly, in 2021 the UK announced its intention to ‘no longer give public figures for our operational stockpile, deployed warhead or deployed missile numbers’. These decisions provide political cover for other nuclear-armed states to reject calls for greater nuclear transparency. They also make it significantly more difficult for external researchers to make accurate nuclear force estimates by denying the public and civil society access to factual information. Such policies also suppress the public’s ability to develop informed opinions about the status and future of nuclear weapons and may serve to facilitate rumours, exaggerations, worst-case planning and fearmongering.

Evaluating information

To produce the annual estimates of world nuclear forces, SIPRI researchers rely on a wide range of publicly available sources, including official and unofficial government statements and intelligence estimates on national or other states’ nuclear weapon arsenals or policies; official documents related to budgets, weapon programmes or nuclear weapon treaties; information from news media outlets and social media; analysis by other think tanks and individual researchers; and commercially available satellite imagery. SIPRI researchers also rely on private conversations with officials, some of whom do not want to be identified, and benefit from occasional leaks from political and military sources.

As noted below, there is a need for critical scrutiny when verifying, prioritizing, combining and using these various information sources. This will differ from country to country, depending on the availability of relevant sources.

Government sources, documents and treaty disclosures

Government documents represent an important official data source. Relevant sources may include budget documents, descriptions of weapon programmes, official press releases, treaty documents, intelligence assessments, declassified documents, testimony to national parliaments, and speeches. Under the best of circumstances, government sources are reliable providers of relatively accurate information; however, in the nuclear context these sources must still be scrutinized and checked for consistency and reliability.

When governments do make statements about either their own or other states’ nuclear arsenals, the statements are often made in a political context or represent the views of a specific government agency rather than the entire national intelligence community. For example, a military agency might make claims about an adversary’s nuclear arsenal as part of an effort to convince lawmakers to increase the agency’s funding. States may also use such statements to signal to adversaries or allies that they have a robust deterrence capability or to communicate that their nuclear policies are responsible or within agreed limits. In addition, different components of a state’s national intelligence community may disagree about the interpretation of available data, or the statements may be worded ambiguously—in some cases probably so as not to reveal too much about the state’s spying activities—leading to possible misunderstanding by the general public. In recent years, for example, US military and intelligence agencies have repeatedly asserted that China will ‘at least double’ the size of its military nuclear stockpile over the next decade. This projection has often been coupled with claims that China plans to alter its long-standing policy of maintaining a minimum nuclear deterrent, or that China might soon adopt a ‘launch-on-warning’ doctrine.[2] It remains to be seen whether this will happen, particularly because many previous projections from the US intelligence community about Chinese nuclear weapons have proven to be exaggerated.

Similarly, while the 2018 US Nuclear Posture Review claimed that Russia had ‘up to 2000’ non-strategic (tactical) nuclear weapons (a claim that has since been regularly repeated by government officials), in 2021 the US Defense Intelligence Agency stated that the figure was 1000–2000, an assessment that would indicate that the number might actually be closer to 1500. Non-strategic nuclear weapon estimates are particularly prone to exaggeration because almost all the related launchers are dual-capable (they can launch both conventional and nuclear warheads). However, an increase in the number of nuclear-capable launchers in a weapon category, does not necessarily mean that the number of nuclear warheads assigned to that category is also increasing. Given the competitive military and political environment in which such statements are being made, it is important not to accept them at face value but instead to analyse them in the context of other information.

Budgetary and programmatic documents often include important details about nuclear weapon systems, such as the number of units produced, their components and subcomponents, their production and deployment locations, and their projected deployment dates. The USA publishes a large amount of such information. The UK, France and Russia publish some information, although Russia publishes considerably less than the other two. China, India and Pakistan publish very little information, while Israel and North Korea generally do not publish any detailed nuclear weapon-related information. Budgetary and programmatic documents can be very useful for tracking and verifying force structures and cross-referencing information.

Treaty-related documents and declassified intelligence assessments can provide a wealth of information about the history and status of nuclear forces. For example, US documents related to the 2010 Russian–US Treaty on Measures for the Further Reduction and Limitation of Strategic Offensive Arms (New START) are a helpful source of disaggregated strategic forces deployment data when assessing Russian and US nuclear forces. Additionally, some documents related to New START and the now defunct 1987 Russian–US Treaty on the Elimination of Intermediate-Range and Shorter-Range Missiles (INF Treaty) provide overviews of nuclear arsenals and base locations that, in many cases, are still valid today. Some of these documents are automatically declassified and made public by the US Government after a certain number of years. In countries with freedom of information laws, such as the USA, large inventories of documents have been declassified and released to the public, and many documents have been published on open-access websites.

Stocks of fissile materials

Information on a nuclear-armed state’s fissile material stocks (i.e. stocks of plutonium and highly enriched uranium, HEU) and its past production of these materials can give some indication about the potential size of the state’s nuclear weapon stockpile. Such information can be derived from official sources such as the International Atomic Energy Agency, from independent estimates by organizations such as the International Panel on Fissile Materials, and from on-site inspections and visits such as Siegfried Hecker’s tours of North Korea’s Yongbyon Nuclear Scientific Research Centre in North Pyongan province.

As with other categories of information, data on fissile material inventories should be used with caution—not least because assumptions about production capacity and efficiency might be too high or too low. A common oversimplification used in the public debate is to calculate the number of nuclear warheads a state might have by dividing the amount of plutonium and HEU that a state has produced with the assumed critical mass of these materials. Fissile material inventories, as well as their management and use in the design of nuclear warheads, vary from state to state. Although knowledge about fissile material inventories helps to set an upper limit for the number of warheads a state can hypothetically produce, SIPRI does not derive its estimates from a simple ‘translation’ of fissile materials into warheads.

Generally speaking, nuclear-armed states do not convert all of their fissile material into weapons; some is kept in reserve. Therefore, in its estimates of the number of warheads produced by a state, SIPRI uses information on fissile materials in combination with state-specific factors such as weapon design and technical know-how, nuclear force structure and nuclear strategy. The assumption that all available material is used for weapons has produced highly inflated warhead estimates in the case of Israel, where it is not uncommon to see estimates of 200–400 warheads (SIPRI estimates that Israel has fewer than 100, based on leaked intelligence estimates and analysis of its nuclear force structure and strategy). The same is true for North Korea, where estimates of 60 or more warheads are common (SIPRI’s estimate is that North Korea could potentially produce fewer than 50 warheads but that the actual assembled number is probably lower, based on analysis by other experts and US statements about North Korea’s missile and warhead capabilities).

Media sources

News media sources regularly report on nuclear weapon developments and play an important role in increasing nuclear transparency and policy accountability. However, the quality of reporting varies considerably between outlets and countries. It is vital therefore to evaluate media reports very carefully and try to trace the story back to the original source. Media embellishment in reporting on nuclear weapon issues is widespread and can often lead to exaggeration and mischaracterization of states’ nuclear strategies and weapon capabilities.

One challenge is that media reports often reproduce what others have written without checking the original source—in some cases without even mentioning any relevant initial media source. For example, several Indian media sources reported in 2020 that one of India’s newest short-range ballistic missiles—the Shaurya—is nuclear-capable. But this has not been publicly confirmed by the Indian Government or by other states in their public assessments of India’s missile arsenal. The claim may have been driven by the fact that the missile is a land-based variant of one of India’s nuclear-capable submarine-launched ballistic missiles (the K-15); however, this does not necessarily indicate that the Shaurya itself has also been assigned a nuclear role.

Another example is the claim in some media that China’s modified Type 094A nuclear-powered ballistic missile submarine carries 16 JL-3 missiles, when in fact it carries 12 less technically advanced JL-2 missiles. Such claims are often copied and repeated by other news media and, in turn, incorporated into other publications, including those produced by think tanks.

In the nuclear context, media reports often refer to unnamed sources such as ‘a naval source’, ‘someone involved in defence planning’ or ‘an industry source’, without giving the reader any sense of whether the source spoke in an official capacity or even has specific insight on the weapon system being described in the news article.

Internet and social media

The internet is awash with information about nuclear weapons. While the internet can be used to circulate important and accurate information quickly, it is also used to spread false information, either deliberately or unintentionally. For this reason, SIPRI uses primary sources whenever they are available.

Social media platforms such as Facebook or Twitter allow users to disseminate information about nuclear weapon issues and developments rapidly to a wide range of people. This can be beneficial at times, as these platforms allow for fast, crowd-sourced analysis to respond to breaking news events such as missile launches or nuclear tests. For issues related to nuclear forces, social media can be useful for detecting developments such as the deployment of a new nuclear launcher or an exercise involving nuclear-capable weapon systems. However, these platforms also tend to amplify sensational information and are breeding grounds for both misinformation and disinformation. Despite these challenges, social media platforms provide a wealth of information about local developments and are monitored by both government intelligence agencies and non-governmental organizations. As with all other sources, social media reports need to be carefully verified against other sources.

Satellite imagery

Commercially available satellite imagery is a particularly valuable resource for analysing nuclear deployments, construction of nuclear facilities and changes to force structures. Freely available lower resolution imagery can be useful for scanning large areas to monitor the construction of buildings and bases, while more expensive higher resolution imagery can be useful for identifying specific buildings, features of bases, and weapon systems. For example, throughout 2020 SIPRI tracked the deployment of Russian intercontinental ballistic missiles at upgraded bases with new security perimeters that could be identified using commercial satellite imagery. SIPRI also used satellite imagery to monitor the construction of missile bases and silos by China.

Satellite imagery analysis is challenging, however, because features may be difficult to interpret accurately and because states may use camouflage and deceptions to confuse adversaries. By combining satellite observations with other sources, it is possible to reduce uncertainty and increase confidence in the observations.


An increase in transparency by all nuclear-armed states would dramatically improve the quality of public debate, and support advocacy and research efforts on nuclear weapon-related issues by civil society and academia. The nuclear policy space is rife with exaggeration, fearmongering, misinformation and disinformation; the best way to overcome these challenges is for countries to be more transparent about their nuclear force structures and postures, and their policies on the potential use of nuclear weapons. Paradoxically, excessive nuclear secrecy may damage security rather than protect it.



[1] ‘Stockpile’ is a term commonly used to refer to the number of nuclear weapons in a state’s military arsenal. A portion of these warheads are assigned to deployed forces and the remainder are held in reserve. The stockpile does not include retired nuclear weapons awaiting dismantlement. SIPRI refers to the total number of nuclear weapons that a state possesses (the military stockpile plus weapons awaiting dismantlement) as the nuclear ‘inventory’.

A state’s nuclear stockpile can be roughly divided into ‘strategic’ and ‘non-strategic’ (tactical) weapons. Strategic nuclear weapons can be defined as those that are designed to be used against an adversary’s cities, military bases, factories and other larger-area targets to damage the adversary’s ability to wage war. There is no universally accepted definition of what constitutes a ‘non-strategic’ or ‘tactical’ nuclear weapon. However, the US Department of Defense describes them as ‘nuclear weapons designed to be used on a battlefield in military situations’. US Office of the Deputy Assistant Secretary of Defense for Nuclear Matters, Nuclear Matters Handbook 2016 (US Department of Defense: Arlington, VA, 2016), p. 17.

[2] Under a launch-on-warning (LOW) strategy, a state indicates that it would launch a retaliatory nuclear strike upon warning of a nuclear attack while the adversary’s missiles are still in the air and before detonation occurs. A launch-under-attack (LOA) strategy, in contrast, involves waiting to launch missiles until after the first detonations have occurred.



Hans M. Kristensen is an Associate Senior Fellow in the SIPRI Weapons of Mass Destruction Programme and Director of the Nuclear Information Project at the Federation of American Scientists (FAS).
Matt Korda is an Associate Researcher in the SIPRI Weapons of Mass Destruction Programme and Research Associate with the Nuclear Information Project at the Federation of American Scientists (FAS).