Table of Contents

Matthew Harries contributed to this chapter.

New START limits the 2,700 or so nuclear warheads deployed on ICBMs and SLBMs—less than 25 percent of Russia and the United States’ combined total inventory of warheads. It also limits the number of heavy bombers, but they are treated as though each carries a single warhead even though they could carry many more. No other warheads—those in storage, those being transported, and those that have been retired and are awaiting dismantlement—are even indirectly accountable. (Table 4 provides an estimate of the Russian and U.S. inventories.)

New START’s traditional design focuses on limiting (most) long-range delivery vehicles and their launchers. In the process, it is both possible and desirable to limit the number of warheads that are physically deployed on (most of) those delivery systems. However, the complexities of extending arms control to any other warheads—even the warheads for heavy bombers that are stored on bases already subject to inspection—have so far prevented it.

Limiting strategic delivery vehicles and launchers has provided—and continues to provide—considerable value. New START and its predecessors have helped to cap and reverse the arms race and to mitigate the uniquely serious escalation risks created by the threats that Russia’s and the United States’ strategic nuclear forces pose to one another. By contrast, with the primary exception of the now-defunct INF Treaty, nonstrategic delivery systems have been left out of arms control. The often-unarticulated justification (correct or not) has been that because these systems generally carry lower-yield warheads and are more suited to tactical missions, such as battlefield use, the risks of further escalation following the first use of NSNWs would be lower than in the case of the first use of strategic weapons.

Table 4: Estimated Warhead Inventories for Russia and the United States
Warhead Type Russia United States
Operational ~4,500 3,750a
Deployed on intercontinental ballistic missiles and sea-launched ballistic missiles 1,410b 1,340b
Warheads for heavy bombers stored at heavy bomber bases ~200 ~300
Other operational warheads ~2,900 ~2,100
Awaiting Dismantlement ~1,800 ~2,000
Total ~6,300 ~5,750
Note: These figures pertain to actual warheads and do not count one warhead for each deployed heavy bomber (as New START does). They are inferred by the authors from the best data available as of October 21, 2021.

a As of September 2020.
b These figures are not known exactly but are much less uncertain than those marked ~.
Sources: U.S. Department of State, Bureau of Arms Control, Verification and Compliance, “Transparency in the U.S. Nuclear Weapons Stockpile,” fact sheet, October 5, 2021,; U.S. Department of State, Bureau of Arms Control, Verification and Compliance, “New START Treaty Aggregate Numbers of Strategic Offensive Arms,” fact sheet, September 28, 2021,; U.S. Department of State, Bureau of Arms Control, Verification and Compliance, “New START Treaty Aggregate Numbers of Strategic Offensive Arms,” fact sheet, December 1, 2020,; Hans M. Kristensen and Matt Korda, “United States Nuclear Weapons, 2021,” Bulletin of the Atomic Scientists 77, no. 1 (2021): 44; and Hans M. Kristensen and Matt Korda, “Russian Nuclear Weapons, 2021,” Bulletin of the Atomic Scientists 77, no. 2 (2021): 91.

Nonetheless, the distinction between strategic and nonstrategic delivery systems is unsatisfactory. Any nuclear detonation would have “strategic” consequences whether the delivery vehicle involved was strategic or nonstrategic. Nominally nonstrategic delivery systems—most notably, SLCMs—have long been able to reach targets deep in an adversary’s homeland and thus threaten that state’s strategic forces and C3I systems. Meanwhile, strategic systems could be used in tactical roles. For example, U.S. Trident D5 SLBMs loaded with low-yield nuclear warheads are technically capable of being used in this way—though the United States asserts that this capability “is not intended to enable, nor does it enable, ‘nuclear war-fighting.’”1

These conceptual flaws have become more glaring in the decades since the Cold War ended. The U.S. Department of Defense assesses that Russia maintains “up to 2,000” nuclear warheads for nonstrategic delivery systems—compared to the estimated 230 in the U.S. arsenal—likely to compensate for its perceived conventional inferiority relative to NATO and possibly China.2 However, the United States and its allies worry that Russia may use this force to enable conventional aggression, particularly against the Baltic states, by seizing territory and then threatening nuclear use should NATO launch a conventional response. If NATO ignored these threats and responded anyway, there is concern that Russia would engage in limited nuclear use to try to coerce NATO into backing down for fear of further escalation.

As a result, the Biden administration has reiterated the long-standing U.S. goal of “new arms control that addresses all of Russia’s nuclear weapons.”3 This is easier said than done. U.S. and NATO concerns are primarily about Russia’s perceived nuclear doctrine—but doctrine cannot credibly be constrained by an agreement. The traditional arms control approach of constraining delivery systems does not seem much more promising because nonstrategic delivery systems, such as tactical aircraft, are all dual-use and neither state has any interest in limiting them.

As a result, the United States has come to focus on the warheads themselves. In the abstract, the United States would probably like to limit only nonstrategic nuclear warheads (that is, the warheads carried by nonstrategic delivery systems). However, nonstrategic warheads in storage cannot be readily distinguished from strategic warheads (those carried by strategic delivery systems)—not least because classification rules would prevent inspectors from directly viewing any warheads. Furthermore, even if discrimination were possible, the difference in size between the Russian and U.S. arsenals of nonstrategic warheads would make reaching an agreement to limit only those warheads politically infeasible. As a result, within the United States, considerable support for a treaty that would impose a single limit on all Russian and U.S. warheads, irrespective of their type or deployment status, has emerged (even if it is not entirely clear that such an agreement would actually address the United States and allies’ concerns).4

Russia has previously expressed support for a limit on all warheads or at least for steps in that direction. In 1997, Moscow and Washington agreed to a framework for negotiations over START III, as part of which they committed to negotiate provisions related to the transparency and destruction of strategic warheads and to explore transparency and confidence-building measures for nonstrategic systems.5 This agreement catalyzed some technical investigations, including on warhead verification, but it petered out after negotiations ended without a treaty.

More recently, Russia has expressed support for a warhead limit as a way to redress the asymmetry between the U.S. and Russian stockpiles of strategic warheads. While most Russian warheads not loaded onto ICBMs or SLBMs are nonstrategic, most U.S. warheads in that category are strategic. In the 2000s, some Russian officials and experts supported a warhead limit on the grounds that it could constrain the United States’ ability to upload reserve strategic warheads onto its larger force of strategic delivery systems. In 2009, for example, Foreign Minister Sergey Lavrov indicated that Russia would seek to use upcoming negotiations over what would become New START to limit “all warheads . . . not only so called operationally deployable, but also the warheads in warehouses.”6 (Given the context for these remarks—an agreement to negotiate a new strategic offensive arms control agreement—it is possible that Lavrov was actually calling for limits on all strategic warheads rather than all warheads, but his concern over the United States’ so-called upload potential was clear.) Somewhat to the surprise of U.S. negotiators, however, Russia did not press this issue during New START negotiations and has not flagged it much since.7 In fact, recently, it is U.S. officials and experts that have begun to express concerns about Russia’s upload potential.8

In late 2020, Moscow and Washington publicly agreed, in principle at least, to a one-year freeze in overall warhead numbers in return for a one-year extension of New START. However, they could not finalize an agreement, largely because of disagreements over verification: Russia refused it, while the United States reportedly insisted on perimeter portal monitoring of assembly/disassembly facilities.9 This impasse does not bode well for Russia’s willingness to negotiate over a verified warhead-limitation treaty.

Even so, given the support that such a treaty attracts in the United States, the concept is analyzed in detail here. Ultimately, even if both sides supported it, verification challenges would likely preclude agreement today. We, therefore, do not try to sketch out a detailed verification protocol, but instead identify critical challenges and a research agenda, comprising unilateral and joint efforts, to solve them. If Russia ever decides to support a warhead limit again, these steps should leave the two states well placed to begin treaty negotiations. But, even if Russia does not, many of the steps would still be useful risk-reduction measures in their own right.

Solution Concept

A warhead limit should apply to all Russian and U.S. warheads regardless of their type, location, or deployment status. A more difficult issue is whether it should also apply to the Russian and U.S. warheads awaiting dismantlement, of which there are currently thousands. Many U.S. experts argue it should not. However, warheads awaiting dismantlement are indistinguishable from operational warheads in storage and are not even stored separately.10 Moreover, by the time a warhead limit could realistically be negotiated, both states would likely have far fewer warheads awaiting dismantlement, making it more feasible to count them under the treaty’s central limit.

One potential complication here is the somewhat different maintenance cycles for Russian and U.S. warheads. Russian warheads have shorter service lives than their U.S. counterparts and must be remanufactured regularly. A 2001 study, for example, estimated that Russian pits (the fissile cores of nuclear weapons) were viable for only seven to fifteen years.11 By contrast, some U.S. warheads were manufactured more than forty years ago.12 Russian warhead service lives have likely increased in the last two decades. Nonetheless, going forward, Russia’s queue of warheads awaiting dismantlement will likely become longer than the United States’ queue. This asymmetry is a potential problem because a treaty that limited the two sides to parity in total inventories would force Russia to retain a smaller operational stockpile, which it could regard as unfair (while the United States would regard unequal limits on total inventories as similarly unfair). However, this issue may not become a significant political challenge if Russia’s dismantlement queue does not become too much bigger than the United States’ queue.13

A treaty would also be more feasible if it did not require Russia or the United States to make major changes to their force structures. The numerical ceiling should therefore require only modest reductions in the U.S. and Russian warhead inventories (as they are constituted at the time of entry into force). This relatively conservative approach is also appropriate given the magnitude of the implementation challenges. Specifically, the size of a militarily significant violation comes down with the warhead limit. Avoiding an overly ambitious target for reductions would therefore reduce the load carried by a largely novel verification regime. If treaty implementation were successful, a successor agreement requiring deeper warhead reductions (along with potentially more intrusive verification measures) could subsequently be negotiated.

As with all arms control agreements, the parties should be given some time to make the required reductions: five years seems appropriate for a treaty that should remain in force for, say, fifteen years. To avoid the accusations of bad faith that occurred early in New START’s implementation, the parties should agree not to increase the net size of their warhead inventories in the meantime.

Central Limits in a U.S.-Russian Warhead-Limitation Treaty

Russia and the United States should agree that each party will reduce and limit its nuclear warheads. The numerical limit chosen should require that each state make modest reductions in its nuclear warheads following the treaty’s entry into force. This limit should apply to all the parties’ warheads, irrespective of their type, location, deployment status, and whether or not they are awaiting dismantlement.

Each party should further commit to reduce its nuclear warheads to meet the treaty’s central limit within five years of entry into force and, in the meantime, to not increase its nuclear warheads above the number it possessed at the time of entry into force. The treaty should remain in force for fifteen years.

For the purposes of treaty implementation, a nuclear warhead should become accountable when fissile material and conventional high explosives or any other energetic material are first assembled within a single device. A nuclear warhead should cease to become accountable as soon as fissile material and conventional high explosives or any other energetic material are no longer assembled within a single device.


Verifying a warhead limit would involve data exchanges, notifications, and on-site inspections, supplemented by each state’s NTM capabilities. These measures build directly on New START’s verification provisions and assume that New START or a successor treaty with similar provisions will remain in force.

Conceptually, warheads can be divided into four categories, depending on how inspections could be most easily facilitated (and, indeed, whether they would be possible at all).

  1. Warheads deployed on ICBMs, SLBMs, and heavy bombers. Under New START, inspectors are already permitted to count the warheads (or, more specifically, the shrouded reentry vehicles) emplaced on deployed ICBMs and SLBMs. Additionally, New START permits inspectors to count any nuclear armaments located on heavy bombers—though, as a matter of policy, neither Russia nor the United States load nuclear weapons onto aircraft in peacetime.14
  2. Nondeployed warheads stored at ICBM, SLBM, or heavy bomber bases. Inspectors already visit these bases pursuant to New START but have no rights to inspect warheads in storage. Nonetheless, it would be sensible to allow verification of those warheads during inspections pursuant to New START or a successor agreement.
  3. Other warheads in storage. Neither Russia nor the United States keeps nonstrategic warheads mated with delivery systems in peacetime. Most, if not all, of these warheads, as well as many reserve strategic warheads and warheads awaiting dismantlement, are stored on bases that are not subject to New START inspections. As a result, a dedicated inspection regime would be needed to verify them. (In a crisis or conflict, some of these warheads could be loaded onto delivery systems, but if that were to occur, treaty implementation would inevitably be suspended.)
  4. Warheads in transport. Warheads are regularly transported, including after assembly and before disassembly. Warheads in transit could not realistically be subject to inspections; rather, the goal would be to use inspections of storage facilities and deployed delivery systems to verify that the consequences of warhead transport operations were consistent with the declarations about those operations.

Data exchanges and notifications. The first step in verification would be a comprehensive warhead data exchange shortly after the treaty’s entry into force (the baseline exchange) and then once every 180 days thereafter. In between comprehensive exchanges, each party would be required to notify the other of any changes to this data and also of any warheads that have been assembled or disassembled. Given the particular security sensitivities around warheads, such notifications should be provided in batches: one exchange every ten days, not more or less continuously as under New START. In addition, each state should be required to provide notifications to the other (1) before the transport of warheads between declared sites, (2) after the transport has arrived at its destination, and (3) after all the warheads in the transport have been unloaded. For security reasons, the notification prior to a transport need only specify a ten-day window in which the transport is expected to depart. (Warheads can also be transported within a site. Although such transports would not need to be declared specifically, the parties would be required to declare the resulting changes to the baseline data.)

Verification: Data Exchanges and Notifications

Within ninety days of the treaty’s entry into force, each party should provide the other with the following baseline information (correct as of noon on the first full day after entry into force):

  1. Its total number of warheads
  2. The number of those warheads in transport between sites
  3. A list of all sites on which its nuclear warheads are located; the number of warheads at each site deployed on ICBMs, SLBMs, or heavy bombers; and the number of warheads at each site not deployed on ICBMs, SLBMs, or heavy bombers
  4. For each such site, a site diagram, conveying inspection boundaries and the locations of any fixed structures wholly or partially within the inspection boundaries
  5. For each such fixed structure, a diagram of its interior, showing all rooms and interior and exterior access points within the inspection boundaries
  6. For each such room, the number of warheads stored within the room; and a list showing the type and UID of, and the number of warheads within, each warhead storage container in that room
  7. For each type of warhead storage container present at any site, its physical dimensions, the maximum number of warheads that it can accommodate, and a photograph

For ICBM, SLBM, and heavy bomber bases, the inspection boundaries should be identical to those declared under New START or any successor treaty, except that they should also include the whole of each fixed structure.

For other sites at which nuclear warheads are present, the inspection boundaries should include the whole of each fixed structure, except that for fixed structures in which warheads are assembled and disassembled, the inspection boundaries should include only those rooms that currently store, are designed to store, or have stored complete nuclear warheads and in which assembly or disassembly does not occur.

Starting 100 days after entry into force and every ten days thereafter, the parties should notify one another of any changes to categories 1, 2, 3, and 6 of the baseline information that occurred during a reporting period that began at noon eleven days earlier and ended at noon one day earlier (except that the first such exchange should cover the period that began at noon on the first full day after entry into force and ended at noon one day earlier). As part of these exchanges, the parties should provide notifications detailing:

  • The total number of warheads that were assembled in the relevant reporting period and the UIDs of the warhead storage containers for those warheads
  • The total number of warheads that were disassembled in the relevant reporting period and the UIDs of the warhead storage containers for those warheads
  • The number of warheads that were removed from warhead storage containers and placed onto ICBMs, SLBMs, or heavy bombers and the UIDs of the relevant warhead storage containers
  • The number of warheads that were removed from ICBMs, SLBMs, or heavy bombers and placed inside warhead storage containers and the UIDs of the relevant warhead storage containers
  • The number of warheads that were moved from one warhead storage container to another and the UIDs of both the new and old containers

Starting 180 days after the treaty’s entry into force and once every 180 days thereafter, each party should provide the other with updated comprehensive data for categories 1, 2, 3, and 6 of the baseline information (current as of noon one day earlier).

The parties should also be required to provide the following notifications concerning the transport of warheads between declared sites:

  • A notification of an upcoming transport that specifies (1) the origin site for the transport, (2) the destination site for the transport, (3) the number of warheads in the transport, (4) the UIDs of the warhead storage containers and the number of warheads within each container, and (5) the first and last day of a window lasting no more than ten days in which the transport is expected to depart.15 This notification must be provided at least one day prior to the first day of the departure window. If the transport does not occur within the ten-day window, the state must provide a cancellation notification within twenty-four hours of the end of the window.
  • A notification that the transport has arrived at the destination site within twenty-four hours of arrival.
  • A notification that all warheads in the transport have been unloaded within twenty-four hours of unloading.

In implementing these provisions, the following definitions would apply:

  • “Fixed structure” means a unique structure that is used to store nuclear warheads, has been used to store nuclear warheads, or is designed to store nuclear warheads.
  • “Room” means an interior subdivision within or underneath a fixed structure that has an access point that is large enough for a warhead storage container to pass through. (The procedure to determine whether an access point is large enough for a warhead storage container to pass through is detailed in chapter 2.)
  • “Noon” means 12:00 in the afternoon in the capital city of the state submitting a declaration or notification.

The inspection boundaries, which demarcate the area to which inspectors have access, are a potential source of controversy here. Under New START, the host state unilaterally imposes those boundaries. This process is uncontroversial because ICBMs, SLBMs, and heavy bombers are large and it would be clear if the host state had drawn the inspection boundaries to exclude, say, some ICBM silos. By contrast, warheads are much smaller and could be located outside of declared storage facilities—creating real potential for disagreement about whether the rules for drawing inspection boundaries had been faithfully followed. A first step toward solving this problem could be to negotiate inspections of empty warhead storage facilities (see chapter 2) as a way of gaining shared understanding about the challenges associated with drawing inspection boundaries and identifying possible solutions.

On-site inspections. Russia and the United States would attempt to verify this data through both on-site inspections and NTM. Inspections, however, would face a fundamental difficulty: the classification rules surrounding warheads would prevent inspectors from viewing them directly (in fact, even reentry vehicles are shrouded during New START inspections). These rules would also prevent inspectors from conducting any measurements that could reveal sensitive design information, which includes, in the case of the United States, “total weights; quantities of contained materials, including but not restricted to tritium, highly enriched uranium, and plutonium; and dimensions, configurations, and weights of fabricated components.”16 Russian rules appear to be similar.

Verifying a warhead-limitation treaty would be easier if these rules were loosened (a notable precedent is the Soviet Union’s permitting independent experts to measure radioactive emissions from a SLCM warhead in 1989—an experiment that would be unthinkable today).17 For this reason, Moscow and Washington should individually assess whether the benefits for warhead verification of revealing additional information to one another would outweigh the security risks of doing so. Nonetheless, the treaty proposed here assumes that neither Russia nor the United States would be willing to change their classification rules.

Because of these rules, verification would largely focus on warhead storage containers on the assumption that, to prevent nuclear accidents, Russian and U.S. warheads are always kept in such containers when not attached to strategic delivery vehicles. Against this background, it is useful to first outline what the inspections would entail and then analyze the challenges created by inspectors’ inability to view warheads directly.

A treaty would not need to contain any specific verification provisions for warheads deployed on ICBMs, SLBMs, or heavy bombers, which are accountable under New START and should be under a successor. However, the agreement would provide for inspections of warhead storage facilities. It should permit those facilities on ICBM, SLBM, or heavy bomber bases to be inspected during the inspections pursuant to New START or a successor and provide for dedicated inspections at other such facilities.

To inspect a storage facility, the inspection team would arrive in the host country at short notice and then designate an inspection site. Almost immediately thereafter, the host state would be required to restrict operations at the designated site to prevent it from hiding evidence of any noncompliance. The design of these restrictions could be controversial. In particular, given the portability of warheads, restrictions on vehicle movements on and out of the base would be helpful but also disruptive (they would prevent security patrols around the site’s perimeter, for example).18

Prior to the start of an inspection, the host state would provide the inspection team with any updated information (since the most recent data exchange) on the location and number of warheads at the site (categories 3 and 6 of the baseline information). The state could also shroud any sensitive objects, other than warhead storage containers, within the inspection boundaries to protect classified information unrelated to treaty implementation.

During an inspection, inspectors would be permitted to count, and read the UIDs of, all storage containers in an agreed number of rooms (this number would need to be calculated using classified information about the design of such facilities to ensure that inspections could produce the requisite level of confidence in the host state’s compliance).19 Inspectors would also verify the number of warheads in any types of containers that could store more than one (see below) as well as the absence of nuclear warheads in shrouded objects and storage containers declared not to contain them (see chapter 2).

Additionally, inspectors should be permitted to visit some number of rooms declared not to contain warheads to verify their absence (many of the procedures set out in chapter 2 would again be useful here). A problem could arise, however, if there were any such rooms that were considered too sensitive to be inspected. Security control centers are one possible example—though Russia has permitted U.S. personnel to visit these areas before to verify the installation of security equipment funded by the United States.20

Inspections would face a fundamental difficulty: the classification rules surrounding warheads would prevent inspectors from viewing them directly.

Russia and the United States should agree to exempt from inspection all warheads in transport (including warheads inside transport vehicles located on declared sites). Instead, they would focus on verifying that the number of warheads and UIDs of storage containers at declared sites after a transport are consistent with the declared details of that transport. They should also agree that the small number of warheads likely undergoing maintenance at declared storage sites should not be subject to inspections.

Inspections would be required, however, at assembly/disassembly facilities. Both Russia and the United States periodically disassemble old nuclear warheads and assemble new ones; Russia probably does so at a significantly higher rate. Therefore, even if an agreement did not require overall reductions in warhead numbers, the parties would still need to be able to remove warheads from accountancy upon dismantlement and introduce new ones after assembly. Verifying such processes—removal in particular—would be challenging.

Warhead verification. In many circumstances, inspectors’ inability to view warheads would not impede effective verification. During a New START inspection of an ICBM or SLBM base, for example, inspectors are permitted to select one deployed delivery vehicle and count the emplaced reentry vehicles, after they have been shrouded, to verify the number declared by the host state. Importantly, the host state would gain no advantage by claiming that an empty reentry vehicle was a warhead. Moreover, it is also physically impossible to place more than one warhead inside a modern reentry vehicle. In consequence, New START inspectors do not need to verify that an object declared to be a warhead is actually a warhead; they simply assume that each reentry vehicle contains one warhead.

For similar reasons, in verifying a warhead limit, inspectors could generally also assume that there was a warhead in any storage container that was declared to contain one. Two important exceptions would arise, however.

First, at least one type of storage container that can hold multiple warheads appears to be in use. Specifically, under the Cooperative Threat Reduction program, Russia was supplied with so-called super containers that can accommodate more than one standard container.21 These super containers were intended to be used primarily for transport, when warheads would not be inspectable, but would inevitably be used at declared sites at other times (including just before and just after transport). If these super containers are still in use—as seems likely—or if any other Russian or U.S. container types can hold more than one warhead, then it would become necessary for inspectors to verify the number of warheads inside them. After all, in this case, a state could gain an advantage by understating this number.

This problem seems manageable. Inspectors should be permitted to request the opening of some number of containers declared to contain more than one warhead to count the number of standard warhead storage containers inside. Although implementing this process might not be entirely straightforward—for example, the lids of the super containers appear to be so large that they probably require a crane to open—there do not appear to be any showstoppers.22

Warhead dismantlement would present a second set of much greater challenges. A party could gain an advantage by falsely declaring that empty storage containers contained warheads awaiting dismantlement (by subsequently “dismantling” those nonexistent warheads, the state could retain more warheads than it had declared). As a result, it would be necessary to verify that an object declared to be a warhead awaiting dismantlement really was an actual warhead. This process would be challenging for two reasons: first, it would be necessary to define what a warhead is in terms of measurable criteria, and second, the unauthorized disclosure of classified information during the measurement process would need to be prevented.

This process is entirely different from verifying that an object declared to be nonnuclear is, in fact, nonnuclear—a verification challenge that arises in New START, for example, because nonnuclear penetration aids (designed to help an incoming missile to defeat missile defenses) may be located on ICBMs and SLBMs alongside reentry vehicles. In this case, inspectors are permitted to test objects declared to be nonnuclear with radiation detection equipment. Because these objects are not radioactive, a null result is expected and there is no danger of classified information being released (unless the inspected party is cheating). By contrast, performing a similar measurement on an actual warhead would likely reveal classified information.

Conceptual solutions to the challenge of warhead verification have been developed.23 Warheads can be defined in terms of their attributes (such as the presence of a minimum quantity of fissile material and the material’s isotopic composition) or their similarity to a template (that is, an object that both sides accept is a real warhead of a given type). To avoid the unauthorized disclosure of classified information, a so-called information barrier would sit between the detection equipment and the inspector to provide an approved output—perhaps just a red light or a green light. Moreover, various approaches have been implemented and tested through both computer modeling and experiments.

This technology, however, is far from ripe for use in treaty verification. Russia and the United States must do more than simply choose hardware, though that could prove difficult enough. They would have to agree on the choice of attributes (which need not be the same for every warhead type) or how to select templates (which would have to vary between warhead types). They would also have to gain confidence that detection systems could not be compromised by the host in an effort to enable cheating (that is, by giving a green light to a fake warhead)—a particular challenge if software is used to convert the outputs of detectors into an unclassified result.

The perquisite to negotiating warhead verification provisions in a treaty is reinvigorated joint research followed by the joint development, testing, and perhaps even manufacture of detection systems.

To be sure, government and independent researchers have tackled these questions from a technical perspective. But, for two reasons, the politics here cannot be ignored: the answers depend on how much confidence each state wants from the verification process, and more importantly, those answers only become solutions when both states agree on them. Today, in the absence of sustained joint research, it seems unlikely that Russia and the United States would have the shared understanding necessary to reach those solutions. Realistically, the perquisite to negotiating warhead verification provisions in a treaty is reinvigorated joint research followed by the joint development, testing, and perhaps even manufacture of detection systems—a degree of cooperation that would not be feasible under current political circumstances.

The Trump administration tried to bypass these challenges in its proposal for perimeter portal monitoring around assembly/disassembly facilities. However, this concept almost certainly reflected what the administration was prepared to accept politically, not the results of a detailed technical assessment. The concept was clearly inspired by the INF Treaty, which provided for perimeter portal monitoring of certain missile production facilities to verify that they were not producing missiles of a prohibited type.24 However, missiles and warheads are different—in terms of their size, the sensitivity surrounding their exact shape, and the way they are packaged for transport. It seems unlikely that the X-ray scanners that the United States used for INF Treaty verification could provide much confidence that real warheads were being brought into assembly/disassembly facilities without also compromising sensitive design information. Indeed, information barrier technology would almost certainly still be needed as part of a perimeter portal monitoring system for warheads, raising all the same challenges discussed above and adding one more: while warheads are in transit, they are shielded by both their storage containers and heavily armored transport vehicles—severely complicating the verification process.25 Of course, some fraction of the warheads entering a facility could be unloaded before verification, but this would create security issues that would be expensive to solve.

In practice, inspectors would likely need access to the warheads awaiting dismantlement inside assembly/disassembly facilities to verify that they really were warheads. The disassembly process itself could not be monitored without revealing extensive information about weapon design. However, warheads entering the process, as well as some of the components and empty storage containers leaving it, could be tracked in order to build additional confidence that those warheads had actually been dismantled.26 Given the exceptional sensitivity of assembly/disassembly facilities, which were not designed to facilitate transparency, designing and implementing a credible verification system could prove challenging—and, once again, joint research would be required prior to negotiations.

Challenge inspections? The verification measures sketched out so far would be intended to ensure that a state had not understated the number of warheads stored at declared sites or the number it had dismantled. However, cheating could also take the form of retaining illicit stockpiles at undeclared sites—though NTM would probably be adequate for verification in this case. Even with a warhead limit in place, Russia and the United States would still retain a few thousand warheads apiece. Illicit stockpiles would have to comprise hundreds of warheads to be militarily significant. The equipment and activities needed to keep this number of warheads safe, secure, and effective would be difficult to hide from NTM and would be particularly noticeable at an undeclared site where nuclear warheads were not supposed to be present. Therefore, so-called challenge inspections to verify the absence of warheads at undeclared sites should not be necessary—which would boost the feasibility of the treaty because challenge inspections could be highly intrusive and difficult to negotiate.

A related difficulty would derive from the possibility of nuclear warheads’ being stored illicitly in vehicles or buildings on a declared site but outside the inspection boundaries. In this case, it might be possible to grant each party some rights to verify such buildings and vehicles using a protocol like the one for verifying the absence of warheads at empty warhead storage facilities (see chapter 2). Indeed, inspecting such empty facilities would build useful experience for designing a warhead-limitation treaty.


Technical feasibility. A treaty to limit all warheads would require substantially more intrusive inspections than either state has accepted before. One illustration of the scale of the challenge is that it took the United States five years, from 1998 to 2003, to negotiate access to Russian warhead storage sites as part of an agreement to upgrade their security. Even so, U.S. inspectors were ultimately not permitted to enter the storage facilities themselves; they could go only as far as the fence that surrounds them.27

Reaching an agreement that allowed the inspection of warhead storage facilities and assembly/disassembly facilities would require access rights to be negotiated in great detail. A useful point of comparison is New START, which, including its protocol and annexes, is 356 pages long (the vast majority of which deals with verification). A warhead-limitation treaty would undoubtedly be even more complicated.

In negotiating such an agreement, Russia and the United States certainly have some relevant experience to build upon—most notably, the declarations and inspections of deployed strategic warheads pursuant to New START. However, inspections in storage facilities and especially assembly/disassembly facilities would create problems that require more unilateral and, more importantly, joint research to solve.

To that end, Russia and the United States should individually undertake the following tasks:

  • Restart national research programs on warhead-level arms control, including on verification technology and inspection procedures.
  • Identify potential changes to classification rules to facilitate warhead-level arms control and assess the benefits and risks of those changes on the assumption that they would be reciprocated by the other party and would apply exclusively to official U.S.-Russian data exchanges and inspection activities.
  • Assess how much confidence NTM could likely provide that the other state had not illicitly retained any militarily significant stockpiles of warheads at undeclared sites in contravention of the warhead-limitation treaty proposed here.

Because these tasks would be undertaken unilaterally and the results could be kept classified, there is no reason why Moscow and Washington could not start on them immediately.

The two states should also undertake the following cooperative efforts, which would realistically require some improvement in their political relationship:

  • Restart joint research on warhead verification, including on information barriers, and commit to the joint development and testing of such technology in the future.
  • Start (or perhaps restart) joint studies on inspections at warhead storage facilities using tabletop exercises and commit to extending such studies to include on-the-ground exercises and assembly/disassembly facilities in the future.
  • Negotiate reciprocal inspections to verify the absence of nuclear warheads at empty actual or suspected warhead storage facilities (see chapter 2).
  • Negotiate a less detailed version of the baseline information exchange outlined above for implementation on a politically binding basis.
  • Consider whether New START’s replacement should limit all warheads located on ICBM, SLBM, and heavy bomber bases, whether deployed or in storage. The advantage of this approach, compared to New START, is that it would limit all warheads deployed, or immediately available for deployment, on strategic delivery systems (not just the warheads located on ICBMs or SLBMs).28 It would also build valuable experience with inspections of storage facilities. The disadvantage is that pursuing such a complex approach would risk derailing negotiations. Russia and the United States should therefore jointly assess its feasibility at the time those negotiations begin.

If U.S.-Russian political relations improve even modestly, the two states should agree to conduct joint research on an unconditional basis, not least as a minimal step to comply with the Nuclear Non-Proliferation Treaty’s disarmament obligations. Other preliminary steps, such as inspections of empty storage facilities, would probably require a quid pro quo of some kind. If the two states could complete the lists of unilateral and cooperative activities, they would be well placed to start negotiations over a warhead limit—at least from a technical perspective.

Political feasibility. For the United States, a limit on all warheads is an attractive concept. It is consistent with the long-standing bipartisan goal of constraining Russia’s large arsenal of nonstrategic nuclear warheads and with the Biden administration’s stated objective of addressing “all of Russia’s nuclear weapons.”

In practice, however, if there were progress in negotiating such an agreement with Russia, some concerns would inevitably emerge. There would likely be some resistance in the United States and in the allied nations that host U.S. nuclear weapons to the unprecedented intrusiveness of—and the potential disruption caused by—inspections inside warhead storage facilities and particularly assembly/disassembly facilities (such concerns probably explain the Trump administration’s focus on perimeter portal monitoring). There would also be domestic criticism over the non-inclusion of China. Indeed, substantial increases in China’s arsenal, or even further evidence that China may be planning such increases, could complicate the political environment for U.S.-Russian arms control long before such increases changed the U.S.-Chinese deterrence equation.

In the final analysis, however, constraining Russia’s nonstrategic weapons is a sufficiently important goal that these concerns would probably not be deal breakers. Indeed, the U.S. Senate might even provide its advice and consent for the ratification of a bilateral warhead-limitation treaty, especially if it did not mandate deep reductions. For the foreseeable future, Russia will continue to have a significantly larger nuclear stockpile than China and will likely remain the pacing nuclear threat. Meanwhile, there are more feasible—though still challenging—approaches to try to manage the buildup in China’s nuclear arsenal, such as the proposed U.S.-Chinese fissile material cut-off agreement (see chapter 4). More generally, given the complexities of warhead-level arms control, it would be preferable to understand and try to solve the challenges on a bilateral U.S.-Russian basis before attempting to negotiate a trilateral treaty.

By contrast, under current circumstances, Russia is unlikely to support a comprehensive and verified warhead-limitation treaty, which would go far beyond the unverified one-year freeze in numbers that Moscow was prepared to accept in 2020. Certainly, Russian concerns about the United States’ upload potential do not appear to provide a strong enough motivation.

One potential Russian concern is the non-inclusion of France and the United Kingdom (though past Russian insistence that those two states must be included in future arms control arrangements has evaporated quickly).29 Another is likely the intrusiveness of inspections. Most fundamentally, however, Moscow views its large stockpile of nonstrategic warheads as a significant advantage that helps to compensate for its perceived conventional inferiority. It is unlikely to limit its warhead stockpile absent significant concessions by the United States.

Moscow has not indicated what the United States would have to trade for a warhead-limitation treaty. In fact, it has not indicated that a trade is even possible. Its long-standing position is that it will not negotiate over NSNWs until the United States has withdrawn all its nuclear warheads to national territory.30 NATO could not accept withdrawal as a perquisite to negotiations, but it should discuss whether withdrawal could be an acceptable outcome of an arms control process that included a warhead-limitation treaty. Such a debate could prove contentious and would need to be carefully managed, including by agreeing at the outset that the alliance would only agree to withdrawal if it could reach consensus.

In this context, conventional arms control could provide important benefits to both NATO and Russia. Russia worries about its conventional inferiority across Europe as a whole, whereas NATO is concerned about its local inferiority in the Baltic region. Measures to address both imbalances—which may require managing the conventional and dual-use missiles previously constrained under the INF Treaty as well as general purpose forces—could help induce Russia to agree to a warhead cap and NATO to agree to the withdrawal of U.S. nuclear weapons from Europe.

A quite different approach would be to trade a warhead-limitation treaty for limits on ballistic missile defenses. Such a deal would not address the reasons why Russia maintains its large arsenal of nonstrategic warheads, but it might well provide more value to Moscow than the withdrawal of U.S. nuclear weapons from Europe and conventional arms control. Of course, U.S.-Russian negotiations over limits on ballistic missile defenses (which would apply equally to both Russian and U.S. capabilities) would be difficult, and the task facing the United States government in building sufficient domestic support to enable ratification even more so. Yet these challenges lay far in the future. For now, the first step is for Russia and the United States, in both official and unofficial dialogues, to creatively explore the potential trade space.


1 U.S. Department of Defense, “Nuclear Posture Review,” 54.

2 Ibid, 53; Kristensen and Korda, “United States Nuclear Weapons, 2021,” 44.

3 For example, U.S. Department of State, “Secretary Blinken’s Call With Russian Foreign Minister Lavrov.”

4 For example, see Steven Pifer, “The Art of Negotiating Non-Strategic Nuclear Weapons,” National Interest, June 4, 2021,; and Frank Miller, “Talking About ‘Strategic Stability,’” Real Clear Defense, July 8, 2021,

5 Arms Control Association, “The START III Framework at a Glance,” fact sheet, January 2003, This agreement followed bilateral talks in 1994–1995 over the transparency and irreversibility of nuclear-weapon reductions. David Hafemeister, “US Nuclear Security Cooperation With Russia and Transparency” in Nicholas Zarimpas, ed., Transparency in Nuclear Warheads and Materials: The Political and Technical Dimensions (Oxford: Oxford University Press, 1993), 95–96.

6 Sergey Lavrov, press conference at the Conference on Disarmament, Geneva, March 7, 2009,

7 The issue is not mentioned, for example, in Rose Gottemoeller, Negotiating the New START Treaty (Amherst, NY: Cambria Press, 2021).

8 For example, Thomas Countryman, “Russia, China, Arms Control, and the Value of New START,” testimony before the House Foreign Relations Committee, Subcommittee on Europe, Eurasia, Energy, and the Environment, July 25, 2019,

9 Ministry of Foreign Affairs of the Russian Federation, “Statement on New START Treaty Extension,” October 20, 2020,; and Michael R. Gordon and Ann M. Simmons, “U.S., Russia Near Deal to Extend Nuclear Treaty and Freeze Warheads for a Year,” Wall Street Journal, October 20, 2020,

10 In the late 1990s, during the implementation of the Cooperative Threat Reduction program, Russian officials informed their U.S. counterparts that warheads awaiting dismantlement were not stored separately or at separate locations from operational warheads. It seems unlikely that this practice has changed since then. Personal communication, William Moon, September 2021. U.S. officials have privately indicated that, in the United States, active warheads and retired warheads awaiting dismantlament are also stored together.

11 Oleg Bukharin, “Downsizing Russia’s Nuclear Warhead Production Infrastructure,” Nonproliferation Review 8, no. 1 (2001): 124.

12 U.S. Department of Energy, National Nuclear Security Administration, “Fiscal Year 2021 Stockpile Stewardship and Management Plan—Biennial Plan Summary,” Report to Congress, December 2020, 2–9,

13 Not least because the United States would probably be unwilling to limit only operational warheads if Russia’s dismantlement queue becomes significantly larger.

14 Protocol to New START, part 5, section VI, para. 12(a).

15 Individual warhead storage containers may be packed inside larger ones for transport (and perhaps at other times too). In this case, notifications should include the UIDs for both the larger outer container and the smaller inner ones.

16 Theodore B. Taylor, “Verified Elimination of Nuclear Warheads,” Science & Global Security 1, nos. 1– 2 (1989): 7. See also Office of Nonproliferation Research and Engineering, “Technology R&D for Arms Control,” NNSA/NN/ACNT-SP01, U.S. Department of Energy, National Nuclear Security Administration, 2001, 2,

17 Hafemeister, “US Nuclear Security Cooperation With Russia and Transparency,” 85.

18 The authors of this report advocate these restrictions as part of the inspection protocol for verifying the absence of warheads at empty warhead storage facilities (see chapter 2). However, there would be far fewer such inspections, and the restrictions would presumably be less problematic at sites that, because they do not contain warheads, require less security.

19 National Academy of Sciences, Monitoring Nuclear Weapons and Nuclear-Explosive Materials: An Assessment of Methods and Capabilities (Washington, DC: National Academies Press, 2005), 65–69, available from

20 William M. Moon, “The Story Behind U.S. Access to Russian Nuclear Warhead Storage Sites,” Stimson Center, February 4, 2021,

21 Joseph P. Harahan, “With Courage and Persistence: Eliminating and Securing Weapons of Mass Destruction With the Nunn-Lugar Cooperative Threat Reduction Programs,” U.S. Department of Defense, Defense Threat Reduction Agency, 265–267,

22 Another challenge would arise if there are any container types that can accommodate multiple warheads that are not also inside their own individual containers. In this case, it would be necessary to shroud the warheads before inspectors counted them.

23 For example, Office of Nonproliferation Research and Engineering, “Technology R&D for Arms Control,” 4–17; Alexander Glaser, Boaz Barak, and Robert J. Goldston, “A Zero-Knowledge Protocol for Nuclear Warhead Verification,” Nature 510, no. 7506 (2014): 497–502; and R. Scott Kemp, Areg Danagoulian, Ruaridh R. Macdonald, and Jayson R. Vavreka, “Physical Cryptographic Verification of Nuclear Warheads,” PNAS 31, no. 113 (2016): 8618–8623.

24 Harahan, On-Site Inspections Under the INF Treaty, chapter 5.

25 On the need to use information barrier technology as part of perimeter portal monitoring, see National Academy of Sciences, Monitoring Nuclear Weapons and Nuclear-Explosive Materials, 76.

26 Eric R. Gerdes, Roger G. Johnston, and James E. Doyle, “A Proposed Approach for Monitoring Nuclear Warhead Dismantlement,” Science and Global Security 9, no. 2, (2001): 121–126; and National Academy of Sciences, Monitoring Nuclear Weapons and Nuclear-Explosive Materials, 76–77.

27 Moon, “The Story Behind U.S. Access to Russian Nuclear Warhead Storage Sites.” 

28 It would also capture any nonstrategic nuclear warheads stored on such bases, which, depending on the number of such warheads, could be an issue for Russia.

29 Gottemoeller, Negotiating the New START Treaty, 33.

30 Ministry of Foreign Affairs of the Russian Federation, “Deputy Foreign Minister Sergey Ryabkov’s Interview With the Newspaper Kommersant,” September 22, 2020, Ryabkov’s comments about negotiations over NSNWs were slightly ambiguous—hence the phrase “appears to be sticking.”