The Nine-State Problem — What the Public Gets Wrong About Nuclear Risk in the Modern Era

The dominant public mental model of nuclear risk remains a Cold War artifact: two superpowers, each capable of annihilating the other, locked in a balance of terror so stable that neither dares move. This model is not merely outdated — it is actively misleading. As of January 2025, nine sovereign states possess nuclear weapons, and for the first time since the Cold War, every single one of them simultaneously strengthened its arsenal in 2024. ✓ Established [1]

The Stockholm International Peace Research Institute's 2025 Yearbook — the most authoritative annual census of global nuclear forces — records a combined global stockpile of approximately 12,241 nuclear warheads as of January 2025, of which 9,614 were operationally available, 3,912 were deployed with operational forces, and roughly 2,100 were sitting on high-alert status on ballistic missiles, ready for rapid launch. ✓ Established [1] SIPRI's accompanying press release warned bluntly that a new nuclear arms race is emerging amid the collapse of arms control regimes. [2]

The classical deterrence framework — developed principally by American theorists including Bernard Brodie, Herman Kahn, and Thomas Schelling between the late 1940s and 1960s — was constructed around the specific geometry of a bilateral, roughly symmetrical standoff between the United States and the Soviet Union. Its core logic, mutual assured destruction, presupposes that both sides possess overwhelming second-strike capability and rational, unified command structures. Neither condition maps cleanly onto a world that now includes India, Pakistan, North Korea, Israel, and a rapidly expanding China.

The Biden administration's 2024 Nuclear Employment Guidance formally acknowledged this transformation, directing for the first time the simultaneous deterrence of China, Russia, and North Korea — a tripartite challenge that classical deterrence theory was never designed to manage. ◈ Strong Evidence [5] The structural problem is compounding: in a multipolar environment, deterrence chains become circular and potentially incoherent. Actions designed to deter one adversary may inadvertently signal aggression to another.

The nine-state nuclear world breaks down as follows, per SIPRI's 2025 data: Russia holds approximately 5,459 warheads; the United States approximately 5,177; China approximately 600; France 290; the United Kingdom 225; India 180; Pakistan 170; Israel approximately 90; and North Korea approximately 50. ✓ Established [2] The aggregate numbers can create a false sense of post-Cold War progress: the global total has indeed declined from a peak of approximately 70,300 warheads during the Cold War. [10] But raw warhead counts are only one dimension of risk — and arguably not the most important one in 2025.

The more consequential transformation is structural: the simultaneous erosion of the nuclear taboo through formalized threshold manipulation, the emergence of cyber vulnerabilities in early-warning and command systems, and the onset of what may be the first true multipolar nuclear arms competition in history. Each of these forces is destabilizing in isolation. In convergence, they create conditions that no established deterrence framework was designed to address.

On November 19, 2024, Vladimir Putin formally approved amendments to Russia's nuclear doctrine. The timing was deliberate: the changes came days after the Biden administration quietly authorized Ukraine to use American-supplied long-range missiles to strike targets inside Russian territory. The amended doctrine signaled a lowered threshold for nuclear use, and it triggered a predictable wave of alarm in Western capitals. What received far less coverage was the substantive expert disagreement about what the amendments actually mean.

The analytical tension here is not merely academic. The two positions carry profoundly different policy implications. If Russia's doctrine is primarily theater, the appropriate response may be calibrated counter-signaling and continued military support for Ukraine. If it reflects genuine operational planning, then the West has been progressively crossing thresholds — Ukraine's strikes on Russian territory, Western intelligence sharing, the supply of long-range missiles — without fully reckoning with the cumulative risk.

There is a further layer of complexity specific to tactical nuclear weapons. The Arms Control Association's November 2025 analysis estimates that Russia has over 1,000 tactical nuclear weapons currently deployed. ✓ Established [12] Russia's so-called 'escalate to de-escalate' concept — the idea that detonating a tactical nuclear weapon in response to a successful NATO conventional attack could coerce war termination — has been part of Western threat assessments for years. But as the Arms Control Association analysis observes, the Ukraine war is forcing intensive re-examination of intra-war deterrence in ways that Cold War theory simply did not anticipate: specifically, what happens when an adversary calls the nuclear bluff repeatedly and survives.

Ukraine's Operation Spyder Web — a series of drone strikes targeting Russian strategic aviation assets deep inside Russian territory — crossed thresholds explicitly set out in the 2024 doctrine without triggering nuclear use. ⚖ Contested [6] The 2025 Defense Intelligence Agency Worldwide Threat Assessment assessed that Russia is 'almost certainly' seeking to avoid direct conflict with NATO because it cannot win a conventional confrontation, and that nuclear use in Ukraine remained 'very unlikely' unless Russia faced an existential threat. ✓ Established [13]

What this body of evidence suggests is not that Russia's nuclear threats are meaningless, but that their meaning is more contextually bounded than doctrine language implies — and that Western desensitization to repeated threshold violations may itself be creating a dangerous dynamic in which the credibility of the next threat is progressively discounted.

No dimension of the contemporary nuclear landscape is more consequential — or more underreported in mainstream coverage — than China's nuclear breakout. For decades, Beijing maintained a policy of 'credible minimum deterrence,' holding a modest arsenal of perhaps 200-300 warheads and officially pledging a no-first-use policy. That posture is now undergoing fundamental transformation.

According to SIPRI's 2025 Yearbook, China's arsenal grew to at least 600 warheads by early 2025, adding approximately 100 new warheads per year since 2023, with approximately 350 new ICBM silos nearing completion. ✓ Established [2] At this rate, SIPRI warns China could match US and Russian ICBM numbers by 2030. Perry World House at the University of Pennsylvania, drawing on US Department of Defense projections, notes that China's trajectory points toward approximately 1,500 warheads by 2035 — a nearly eightfold increase from its 2018 baseline of roughly 200. ✓ Established [5]

The silo construction programme is particularly revealing. The United States and Russia each currently field ICBMs in the hundreds; China's 350-silo programme, if fully populated with warheads, would bring it into rough parity on land-based ballistic missiles. Whether Beijing intends to fill every silo — or is constructing a 'shell game' to complicate adversary targeting — remains genuinely contested among Western analysts.

The strategic ripple effects extend beyond the US-China dyad. India's response has been direct and measurable: in March 2024, India tested the Agni-V ballistic missile with Multiple Independently targetable Re-entry Vehicle (MIRV) technology — a capability that signals a shift away from 'credible minimum deterrence' toward potential first-strike capability against both Pakistan and China. ✓ Established [5] India also commissioned its second nuclear-powered submarine in 2024, reinforcing a survivable sea-based deterrent. The India-Pakistan-China triangle is thus undergoing simultaneous doctrinal revision by all three parties — a trilateral action-reaction dynamic for which no existing arms control framework provides any constraint.

The American response to China's expansion has been quantitative as well as qualitative. The Stimson Center's March 2025 analysis notes that the United States plans to produce 80 new plutonium pits per year by 2030, representing the first new warhead design programme since the end of the Cold War. ✓ Established [7] The feedback loop between Chinese expansion, American modernisation, and Indian capability development is the closest analogue the world has yet produced to a genuine multipolar nuclear arms race — and it is operating entirely outside any treaty framework.

In the formal language of strategic studies, the 'stability-instability paradox' holds that nuclear weapons, by raising the stakes of escalation to catastrophic levels, create strategic-level stability while simultaneously enabling — not deterring — conventional and sub-threshold conflict at lower levels. Both parties to a nuclear standoff, knowing that neither can afford to allow a conflict to go nuclear, have strong incentives to fight aggressively below that threshold. The Ukraine war has provided a sustained empirical test of this proposition, and the evidence is striking.

Russia invaded Ukraine in February 2022 shielded by its nuclear arsenal. The implicit — and occasionally explicit — nuclear threat served its primary intended purpose: it deterred NATO from direct military intervention. But it conspicuously failed to deter Western military aid to Ukraine, intelligence sharing, economic sanctions, or Ukraine's own escalating conventional operations against Russian territory. ◈ Strong Evidence [6] Nuclear weapons, it turns out, cannot deter everything — and the attempt to extend deterrence to sub-threshold activities creates a credibility problem: threaten too broadly, and the threat loses its force.

The May 2025 India-Pakistan conflict — involving Indian strikes under what New Delhi called Operation Sindoor — provided a second, even more acute real-world data point. SIPRI's 2025 Yearbook recorded that the early 2025 conflict involved strikes near nuclear-related military infrastructure, with SIPRI warning that third-party disinformation 'risked turning a conventional conflict into a nuclear crisis.' ✓ Established [2] The Global Security Review's October 2025 analysis observed that Operation Sindoor similarly demonstrated the limits of Pakistan's nuclear deterrence against Indian conventional force — Pakistan's arsenal did not prevent Indian strikes, just as Russia's arsenal did not prevent Ukrainian strikes on Russian soil. [6]

What these cases collectively demonstrate is that nuclear weapons in the modern era function less as an absolute prohibitor of conflict and more as a ceiling — one that can be approached from below with increasing confidence as each sub-threshold violation goes unpunished. The danger is precisely this progressive desensitization: each successful conventional operation inside a nuclear state's declared red lines normalizes the next one, compressing the psychological and operational distance between sub-threshold conflict and nuclear use.

Classical deterrence assumes that the threat of retaliation is credible, proportionate, and controllable. Each of these assumptions is increasingly challenged by the intersection of nuclear systems with cyber vulnerabilities. A 2025 study published in the journal Risk Analysis by Zaidi et al. — covering nuclear command systems and power plants across multiple states — found that these facilities are increasingly vulnerable to cyberattacks that can disrupt operations and undermine deterrence. Critically, the study identified that cyber intrusions risk eroding second-strike credibility and may create incentives for preemptive action — the precise opposite of deterrence stability. ◈ Strong Evidence [8]

The study further found that implementation of cybersecurity guidelines remains inconsistent across nuclear-armed states — a finding that translates directly into asymmetric vulnerability. A state confident in the security of its own command systems but uncertain about an adversary's cyber defences may calculate that a crisis window exists in which to act before the adversary's systems can respond. This is the cyber-nuclear entanglement problem in its sharpest form: it doesn't require a successful cyberattack to be destabilising. The mere possibility of one can alter crisis behaviour.

The DEFCON Warning System's December 2025 analysis adds a further dimension: even air-gapped US nuclear command-and-control networks — physically isolated from the public internet — are not fully immune to cyber intrusion. A breach, the analysis notes, could spoof false warning data or inject fake launch signals into the decision chain. ◈ Strong Evidence [9] This is not a theoretical concern: the history of nuclear close calls is replete with cases where false warning data — a flock of geese misread as ICBMs, a Norwegian weather satellite mistaken for a missile — came disturbingly close to triggering inadvertent launches. The question is whether adversarial actors can now engineer such false warnings deliberately.

The Carnegie Corporation of New York's 2025 analysis adds a dimension that has received almost no mainstream coverage: US intelligence has assessed that Russia is developing a space-based nuclear weapon that could threaten all satellites simultaneously. ◈ Strong Evidence [10] Modern nuclear deterrence is critically dependent on satellite infrastructure for early warning, communications, and targeting. A weapon capable of blinding all satellites simultaneously would not merely degrade deterrence — it would potentially render it inoperable in the critical early minutes of a crisis.

Among the least reported nuclear risk factors of 2025 is the profound asymmetry in early-warning capabilities across the nine nuclear states. The United States and Russia both operate mature, multi-layered early warning architectures combining space-based infrared sensors, ground-based radar networks, and hardened command communications. The other seven nuclear states operate in various states of incompleteness — and for some, the gaps are extreme.

The DEFCON Warning System's December 2025 technical analysis describes China's early-warning satellite constellation as the 'largest gap' in its strategic forces, noting that China potentially lacks full global coverage and that its capabilities are largely untested in real crisis conditions. ◈ Strong Evidence [9] For a state adding approximately 100 warheads per year and constructing 350 new ICBM silos, this gap between offensive capability and defensive awareness is structurally dangerous.

Pakistan's situation is more acute still. The DEFCON Warning System analysis finds that Pakistan lacks early-warning satellites entirely, making it the nuclear state with the most compressed decision timelines of any in the world. ◈ Strong Evidence [9] Pakistan's approximately 170 warheads — per SIPRI 2025, all targeted at India — constitute a deterrent premised on rapid response. Without satellite-based early warning, that rapid response must be triggered by ground radar alone, leaving minimal time for human verification before a launch decision must be made.

The implications for the India-Pakistan dynamic are stark. SIPRI's 2025 Yearbook documented an early 2025 armed conflict — Operation Sindoor — in which Indian strikes came near nuclear-related military infrastructure, with SIPRI warning that third-party disinformation risked converting the conventional conflict into a nuclear crisis. [2] Pakistan's compressed decision timeline means that in a future crisis, a false or manipulated warning — whether from faulty sensors, disinformation, or a deliberate cyberattack — could arrive and demand a launch decision within minutes, with no satellite-based confirmation available.

The India-Pakistan corridor thus represents the intersection of three risk factors simultaneously: minimal warning time, cyberattack vulnerability, and active disinformation from third parties. Classical deterrence theory's assumption of rational, well-informed decision-makers with adequate time to deliberate does not describe this environment.

In 2025 Senate testimony, General Anthony Cotton, commander of US Strategic Command, disclosed that the US is incorporating AI into nuclear command-and-control systems with the stated goal of 'accelerating human decision-making.' ✓ Established [10] The Carnegie Corporation's analysis of this testimony notes that senior military officials have acknowledged they cannot predict how AI integration will complicate the system — a remarkable admission for a domain where miscalculation carries civilisational consequences.

The stated rationale is understandable within its own logic: the same compressed decision timelines created by adversary hypersonic missiles and forward-deployed submarines that plague Pakistani planners also affect American ones. If an adversary can design a strike to arrive with only five minutes of warning, AI-augmented decision support may genuinely save minutes that matter. But the introduction of AI into this chain creates new failure modes that have no Cold War precedent.

The Cyber Resilience and Strategic Stability study published in Risk Analysis in October 2025 identifies a critical structural concern: AI systems trained on historical data may perform unpredictably in genuinely novel crisis scenarios — which is precisely the environment in which nuclear decisions must be made. [8] An AI system that 'accelerates' a human commander toward a launch decision based on pattern-matching with historical scenarios could be catastrophically wrong in a scenario that has no historical precedent — which, in a multipolar nuclear crisis involving three simultaneous adversaries, cyber disruption of early-warning systems, and real-time disinformation, is precisely the scenario that must be planned for.

There is also an adversarial dimension. An AI system integrated into nuclear C2 creates a new target for adversary cyber operations. Rather than attacking the physical infrastructure of missiles or command bunkers — hardened and defended at enormous cost — a sophisticated adversary might instead attempt to corrupt the AI's decision inputs, poisoning its training data or injecting anomalous signals designed to produce false confidence or false alarm. This attack vector did not exist in the Cold War and remains almost entirely unaddressed in public arms control discourse.

In 2024, the University of Oklahoma's Institute for Public Policy Research and Analysis conducted the NS2024 Survey on American public knowledge of nuclear weapons policy. The results are striking in their directness: only 30% of Americans report being at least somewhat familiar with US nuclear weapons policy; only 20% are familiar with its costs; and fewer than 40% are familiar with the concept of nuclear deterrence itself. ✓ Established [14]

Yet 63% of those same Americans credit US nuclear weapons as effective at preventing conflict — suggesting that confidence in the system significantly outstrips understanding of it. [14] This asymmetry — high confidence, low knowledge — is precisely the condition under which democratic publics are most susceptible to both alarmism and complacency, depending on which way the information environment tilts.

The Vienna Centre for Disarmament and Non-Proliferation's April 2025 survey — conducted by researcher Herzog across 27,000 respondents in 24 countries — reveals a more complex picture that the pollsters have termed 'strategic morality': views on nuclear weapons that are deeply context-dependent rather than binary. ✓ Established [11] Sixty-five percent of global respondents considered nuclear use morally unjustifiable — and yet large majorities in the same survey said they would support their country using nuclear weapons in response to an attack on an ally. Between 68% and 85% supported joining the Treaty on Prohibition of Nuclear Weapons, yet only 31% supported unilateral US nuclear reductions. [11]

These are not simply contradictory views — they reflect a coherent if under-theorised moral intuition: that nuclear weapons are uniquely terrible in the abstract, while remaining acceptable as a last-resort defence of genuine existential interests. The problem is that this intuition is not equipped to evaluate the specific risk dynamics explored in this report: the compression of decision timelines, the cyber-nuclear entanglement, the multipolar deterrence chains, the tactical weapons trap. Informed public opinion on these questions requires a level of technical and strategic literacy that current educational and media ecosystems are not providing.

The resource dimension compounds this problem. The University of Chicago's Existential Risk Lab estimates that only approximately $40–50 million per year is invested by philanthropy globally in nuclear risk reduction — compared to $91.3 billion in government spending on nuclear weapons in 2023. ✓ Established [15] Even this thin philanthropic base is shrinking: the MacArthur Foundation, which previously provided approximately 30% of philanthropic funding for nuclear risk reduction work, exited the field entirely in 2024. The research, education, and public communication infrastructure needed to close the expert-public knowledge gap is thus operating on a funding base that is, by any reasonable measure, inadequate to the task. [15]

New START — the last remaining bilateral arms control treaty between the United States and Russia, limiting each side to 1,550 deployed strategic nuclear warheads — expired in February 2026 without renewal after Russia suspended its participation in February 2023. There is currently no legally binding agreement limiting the world's two largest nuclear arsenals for the first time since the era of détente. SIPRI's 2025 Yearbook identifies this collapse of arms control regimes as a central driver of the emerging nuclear arms race. ✓ Established [2]

The arms control vacuum is not merely a bilateral US-Russia problem. The entire architecture of nuclear restraint has been progressively dismantled over the preceding decade: the Intermediate-Range Nuclear Forces Treaty was terminated by the United States in August 2019 after US and NATO governments concluded Russia had been violating it for years; the Open Skies Treaty was exited by the United States in 2020 and Russia in 2021; the Anti-Ballistic Missile Treaty had already been withdrawn from by the United States in 2002. What remains is the Nuclear Non-Proliferation Treaty — which China, India, Pakistan, Israel, and North Korea are not party to in the same capacity — and a series of bilateral and multilateral confidence-building measures that were designed for a different threat environment.

The conditions for negotiating replacement agreements are arguably the worst since the immediate post-war period. Russia is engaged in active conflict with a state that previously surrendered its nuclear arsenal on the basis of security guarantees that proved worthless. China has historically declined to participate in bilateral US-Russia arms control frameworks and has shown no indication of willingness to do so as it expands its arsenal. North Korea has developed, per the DIA's 2025 Worldwide Threat Assessment, an ICBM capable of ranging the continental United States — while refusing any engagement with the international non-proliferation regime. ✓ Established [13]

The Stimson Center's March 2025 analysis captures the core policy dilemma: many foundational beliefs about deterrence and nuclear-armed nations' commitments are being reconsidered simultaneously, yet the institutional and intellectual infrastructure for that reconsideration is operating on minimal resources. Defense companies spent $117 million lobbying in 2020, earning $236 in nuclear contracts per dollar spent. ◈ Strong Evidence [15] The asymmetry between the resources devoted to building and maintaining nuclear arsenals and those devoted to understanding and reducing the risks they create is not merely an academic concern — it reflects a structural bias in the political economy of nuclear policy toward procurement and against analysis.

What this report's evidence collectively suggests is that the world is not facing a single, well-defined nuclear threat that existing theory and policy can manage within its existing parameters. It is facing the simultaneous convergence of multipolar deterrence complexity, formalized threshold erosion, cyber entanglement of command systems, compressed and satellite-blind decision timelines for smaller nuclear states, AI integration into the decision chain, and the near-total collapse of the treaty architecture that provided at minimum a shared vocabulary for managing these risks. Classical deterrence theory was a remarkable intellectual achievement for the world it was built to address. It was not built for this one. The most dangerous misconception of the modern nuclear era is the belief that it still is.