Lorenzo Maria Pacini
Classical nuclear deterrence is dead. New START expired, tripolarity returns, verification gone. Iran proves nonproliferation is a trap.
A contingency mistaken for law
We have often discussed nuclear deterrence and how the doctrine has changed, especially in the context of hybrid wars and the new weapons now available. Precisely on this point, at the 23rd annual meeting of the Valdai Discussion Club, my friend Nelson Wong put forward a thesis that deserves attention precisely because it is presented without fanfare.
The classic theory of nuclear deterrence, developed during the Cold War around the concept of mutually assured destruction, was based on historical conditions that have largely ceased to exist today. The framework isn't entirely wrong: mutual vulnerability remains, and the threat of unacceptable retaliation is perhaps still the only thing preventing a direct war between major powers. But too many features of the landscape have changed for that map to serve as a guide. My argument is that the individual factors eroding deterrence share a common root: the transition from unipolarity to a tripolar system, managed by elites trained in an era when competition among equals had been declared over.
Classical deterrence arose from specific circumstances: two main actors, fairly predictable escalation ladders, and functioning channels of communication. That shared grammar was not a given, but the product of costly learning. Thomas Schelling described the diplomacy of violence as a tacit bargaining process in which a threat is valued for what it communicates; the code of bipolarism was written through crises in which the two sides teetered on the brink and drew institutional consequences from them, from the 1963 hotline to the 1972 ABM Treaty. Atlantic simulations confirm this. Carte Blanche (1955) and Proud Prophet (1983) showed that stopping a Warsaw Pact offensive with tactical weapons would have killed about 1.7 million Germans and that keeping the war at the "tactical" level was an illusion, because the leap to the strategic level seemed almost automatic. Deterrence worked in part because those who planned it knew that once the threshold was crossed, it would no longer work. It was a culture of limits rather than a theory of calculation.
The guardrail has shattered
In February 2026, New START expired, and for the first time in over half a century, Russia and the United States are not bound by any binding bilateral nuclear treaty; verification, notifications, and data exchanges have ceased. That same year, the NPT Review Conference ended without results. Verification was not merely a technical mechanism: it was the arena where two military bureaucracies learned to understand one another. With that channel closed, each side has reverted to assessing the other based on worst-case scenarios - the very raw material of the security dilemma.
Added to this is the end of bipolarity. In the realist interpretation that John Mearsheimer has revived with Glenn Diesen, the current disorder stems from the transition toward multipolarity and the persistence of problems inherited from the unipolar era. The liberal order was the global projection of the Western order of the Cold War, possible only as long as the sole remaining great power was a liberal state. Today there are three powers, each with a different conception of order. A tripolar system is not simply a bipolar system with an extra player: with two actors, it is sufficient for one dyad to be stable; with three, all must be stable, and each measures its own arsenal against the potential combined strength of the other two. This is why the growth of China's arsenal weighs on U.S. planning, regardless of Beijing's intentions.
Thus, the security dilemma returns to the forefront: China's intercontinental missiles are defensive for Beijing and offensive for Washington; Taiwan's weapons are defensive for the United States and offensive for China. There is also a more troubling element. The stated goal of inflicting a "strategic defeat" on Russia is not containment, but rollback. During the Cold War, deterrence preserved the status quo; today, the West invokes it while it is itself the revisionist pole, pushing the line forward through NATO expansion and increasingly advanced weapons. This amounts to asking the adversary not to react while one climbs the ladder of escalation. Deterrence that masks a shift in the strategic balance ceases to convey stability and begins to convey an intention.
The Iranian test case
No theater puts these theories to the test quite like Iran. On June 13, 2025, Israel struck the Islamic Republic with Operation Rising Lion; Tehran responded with "True Promise 3." The paradox was already evident: Iran had signed the NPT, and the Supreme Leader had issued a fatwa against nuclear weapons, while Israel possesses an undeclared arsenal estimated at between 75 and 200 warheads and keeps Dimona off-limits to inspections. I wrote at the time that the attack would not destroy Iran's program but would push Tehran to view the weapon as a deterrent. The nonproliferation regime promises security to those who renounce nuclear weapons; the Iranian case teaches us that those without them are targeted, while those who possess them are not.
In the summer of 2026, there was open discussion of using tactical nuclear weapons against a non-nuclear country, in an air and naval conflict whose stated aim was to reopen the Strait of Hormuz. Against whom ? Striking the strait would contaminate the waters of the United Arab Emirates, Kuwait, and Saudi Arabia; striking the interior would not guarantee surrender. In either case, the nuclear taboo would be broken, setting a precedent for Moscow, Beijing, and Tel Aviv; it is not surprising that, according to available information, Trump's advisors oppose this. Nina Tannenwald has shown that the non-use of nuclear weapons since 1945 rests on an established norm, not merely on the calculation of retaliation. Today, that norm holds more because of the prudence of certain advisors than because of its own strength, and a deterrent that depends on the character of individuals is no longer an institution.
The Gulf also shows that the logic of "unacceptable punishment" is shifting toward non-nuclear infrastructure. After the refineries in Tehran, Haifa and the Bapco refineries in Bahrain were struck; and the attack on the desalination plants in Qeshm has effectively legitimized retaliation against Israel's desalination network - on which over 60 percent of the country's water supply depends - and against that of the Gulf, where some states derive 80-90 percent of their drinking water. It is an energy-humanitarian "Samson option": the threshold of unacceptable punishment, which during the Cold War coincided with the nuclear threshold, is now reached by drones launched against a desalination plant. Herman Kahn's escalation ladder assumed visible and costly steps to climb, but in the Gulf, the steps have been crushed one on top of the other.
The material basis and the fearless machine
Deterrence is also a function of the ability to sustain a war over time. While Washington was burning through interceptors over Iran, Beijing was presenting its 15th Five-Year Plan, featuring "extraordinary measures" for self-sufficiency in critical supply chains. China controls about 90 percent of the world's rare-earth refining, on which every F-35, every Patriot battery, and every THAAD interceptor depends; starting in 2025, it will subject these materials to export controls, while the Pentagon must eliminate its dependence by January 2027. If the replenishment of conventional deterrence depends on the rival, the threshold beyond which one is tempted to compensate with nuclear weapons is lowered: a deterrent that cannot be replenished tends to become nuclear out of necessity. Whoever controls weapons materials controls the pace of others' wars.
Then there is artificial intelligence. In 1983, Stanislav Petrov deemed a report of a U.S. launch to be a false alarm; when integrated into command systems, AI can reduce that pause from minutes to seconds, and a system that feels no fear cannot be dissuaded. The immediate danger is not the machine that decides to launch, but the decision-support system that accelerates escalation or excludes de-escalation from the options. I'd like to add a point about the education of decision-makers: if Mearsheimer is correct, the current ruling class came of age after 1991, when the study of the balance of power had all but disappeared from European universities. In 1983, the machine made a mistake, and the human knew it. Today, the risk is that the machine will make a mistake and the human will lack the tools to detect it.
Legitimacy eroded from the center
The NPT established a two-tier system with clear colonial legacies, and the 2017 Prohibition Treaty challenges its legitimacy. But the erosion of the legal framework does not originate at the periphery. It starts at the center. Washington has claimed control over an international strait as national territory and the right to seize Iranian oil tankers; the European Commission has proposed a "reparations loan" of up to 165 billion euros guaranteed by Russian funds frozen at Euroclear. However one judges it, the inviolability of sovereign assets is being called into question by the very states that established it. If sovereign property, freedom of navigation, and nonproliferation are suspended when it suits those who wrote the rules, others draw the most rational conclusion: the only reliable guarantee is one's own strength.
This does not mean that deterrence is dead. Following the August 26 Storm Shadow attack on Donetsk, Maria Zakharova warned that Moscow could strike British installations "in Ukraine and beyond its borders": the most explicit threat to date directed at a NATO member, one to be taken seriously without overreacting, as it remains merely verbal. Ratcliffe's mission to Moscow on August 25 shows that the channels of communication are not closed. The warning system still works; what is missing is the predictability of the response. Deterrence is being practiced, but blindly.
For the 2026-2028 period, based on evidence of moderate quality, the most likely scenario (approximately 55 percent) is unregulated deterrence: no successor to New START, recurring crises managed through unilateral signals and ambiguous thresholds. A minimal restoration (about 30 percent) would establish crisis communication channels, launch notifications, and commitments to human control over nuclear decisions, following the statements made by the United States and China in Lima in November 2024; the indicators would be the resumption of Russian-American consultations and a negotiated formula for the Strait of Hormuz. Breaking the taboo (about 15 percent) - limited use against a non-nuclear state - would be preceded by deployments of low-yield warheads, rhetoric normalizing tactical use, or the depletion of interceptor stockpiles. Everything hinges on one assumption: that no power considers the stakes in peripheral theaters to be existential. A crisis over Taiwan superimposed on the Gulf crisis would bring this assumption crashing down.
Rebuilding crisis communication channels, limiting the integration of AI and nuclear weapons, and giving a voice to the Global South: these are necessary measures, but they are insufficient. Bipolar deterrence was stable because it protected an order that the superpowers - while ideologically opposing it - recognized in practice. Multipolar deterrence requires something analogous: namely, the mutual recognition of security spaces among poles that do not share the same conception of order.
This is not a moral concession but a technical necessity: without a recognized framework, deterrence does not know what to protect. A new map is drawn starting from the territory, and today's territory consists of three poles: mining supply chains, desalination plants, algorithms, and rules applied intermittently. Understanding this is the minimum requirement - and I emphasize minimum ! - for clear-headed thinking to prevail over reassuring formulas.