The Grid is the Kill Web: Rethinking Power Wars as a Mesh Force Problem
Morgan Bazilian and Jahara Matisek have done the defense community a service with their CIMSEC piece, “The Grid is the Arsenal: Power Wars and the New Foundations of Military Strength.” Their argument is blunt and, I think, correct: electricity is becoming to 21st century militaries what oil was to the 20th, and the United States is entering this competition with an electrical grid that was never designed to be a strategic asset.
Where I want to take their argument further is in showing why this is not simply an energy policy problem or an industrial base problem. It is a kill web problem. The grid is not merely the thing that powers the arsenal; it is now an integral, load-bearing member of the kill web itself, and needs to be understood and defended as such.
For several years, Ed Timperlake and I have argued that the central organizing distinction in modern warfare is between the kill chain and the kill web. The kill chain is linear and sequential: find, fix, track, target, engage, assess. It presumes a relatively stable, centralized architecture in which sensors feed a small number of decision nodes that in turn cue a small number of shooters. The kill web is different in kind, not just in degree. It is a distributed, reconfigurable mesh of sensors, shooters, and decision nodes in which any node can, in principle, pass targeting-quality data to any other node, and in which the loss of any single node does not collapse the system.
Joint All-Domain Operations, the concept Bazilian and Matisek rightly identify as the Pentagon’s overarching bet, is simply the kill web made doctrine. It cannot exist as a linear chain; it has to be a mesh, because that is the only architecture that survives contact with a peer adversary who is actively trying to blind and disconnect it.
What Bazilian and Matisek have identified, without quite naming it this way, is that the kill web has a physical substrate, and that substrate is the electrical grid. Sensor fusion, AI-enabled targeting, continuous inference at the tactical edge, secure and redundant communications, the data centers that train and run the models the Pentagon leaned on during the Iran war — none of this exists without an unbroken, high-reliability supply of electrical power.
The authors are correct that this power dependency operates on two very different clocks: the hour-by-hour clock of peak load that determines whether a radar network or a command node stays lit, and the multi-year clock of siting, financing, and building the generation and transmission that stands behind it.
But I would put a finer point on it. What they are really describing is that the American kill web has an unrecognized single point of failure sitting underneath it: a fragmented, balkanized, permission-choked grid that was engineered for 20th century load patterns, not for the always-on, geographically dispersed, computationally hungry architecture that JADO requires.
This is where the second organizing framework in my own work becomes directly relevant: the distinction between intelligent mass and exquisite scarcity. For two decades, American acquisition strategy has chased exquisite scarcity — small numbers of extraordinarily capable, extraordinarily expensive platforms and munitions, optimized for a kind of warfare in which the United States would always have qualitative overmatch and would rarely need to replace what it expended. Ukraine, and now the 2026 Iran war that Bazilian and Matisek cite, have exposed the bill that comes due when that assumption fails. Their arithmetic on interceptor expenditure versus replacement timelines, five weeks of munitions use requiring up to five years to replace, is the exquisite scarcity problem stated in its starkest form.
What their article adds, and what I think deserves to be central rather than incidental to the argument, is that the bottleneck upstream of “we don’t have enough missiles” is “we don’t have enough reliable, surge-capable electricity to make more of them, on the timeline a war actually demands.” Intelligent mass — the capacity to generate large numbers of good-enough systems rapidly and to regenerate combat power under attrition — is not a magazine depth problem or a shipyard problem alone.
It is fundamentally a power problem. An ammunition plant sitting in a power-constrained grid region cannot surge, no matter how much money Congress appropriates for it. Bazilian and Matisek’s point about copper is the same argument at the level of the periodic table: eighteen to thirty years to bring a new domestic mine online means that whatever the United States decides today about its industrial base, it has already decided its constraints for the next generation. Intelligent mass, in other words, has a lead time measured in decades before the first artillery shell is ever forged.
There is a third framework that this article speaks to directly, and it is one I have used to describe the difference between how militaries plan for war and how they actually have to fight it: chaos management versus crisis management. Crisis management assumes a single, bounded, sequential problem that can be diagnosed and resolved by a centralized decision-making apparatus before the next crisis arrives. Chaos management assumes simultaneous, overlapping, non-sequential disruptions across multiple domains at once, in which no single node of authority has a complete picture and the system has to be resilient by design rather than rescued by decision after the fact.
Bazilian and Matisek’s description of the American grid, regional interconnects like ERCOT and PJM that cannot easily move power across their own seams, permitting queues measured in years, transformer lead times that have gone from months to years, is a textbook crisis-management architecture trying to survive in a chaos-management world. China’s answer, the thirty-thousand-kilometer ultra-high-voltage transmission buildout the authors cite, is not simply bigger; it is architected differently, to move surplus power from the interior to the coast on a national, centrally directed basis.
That is a chaos-management-capable grid built by a chaos-management-capable state. The American grid, by contrast, still behaves as though the worst case it needs to survive is a heat wave in one region at a time, not a coordinated adversary campaign against critical infrastructure of the kind Russia has run against Ukraine’s grid since 2022, or the kind that a Taiwan contingency or a Guam contingency would almost certainly visit on the United States and its allies in the Indo-Pacific.
That regional and allied dimension is where I would push the authors’ otherwise domestically-focused argument outward. Their one throwaway line about Guam that its power grid is “falling critically short, putting U.S. bases there at risk and undermining military readiness in the Pacific” deserves to be the lead, not a footnote. I have spent a great deal of time over the past year with the Sir Richard Williams Foundation looking at what “fight tonight” actually requires of the joint and combined force in the Indo-Pacific, and at what mesh forces and maritime ISR architectures need from their supporting infrastructure.
The uncomfortable truth that keeps surfacing in that work is the same one Bazilian and Matisek are describing for the continental United States: distributed lethality and distributed sensing are only as resilient as the distributed power that sustains them. A mesh force concept that scatters sensors, shooters, and unmanned systems across a contested littoral in order to complicate an adversary’s targeting problem is a kill web solution to a kill web threat.
But if every node in that mesh, and every shore-based facility that supports it, draws on a handful of centralized, undefended, import-dependent power sources, the adversary does not need to defeat the mesh directly. It only needs to do to Guam, or to a forward operating base in the Philippines, or to an Australian port, what Russia has done to Ukraine’s grid: attack the substations and transformers rather than the platforms, and watch the entire kill web go dark from underneath. This is precisely the kind of degrade-the-foundation-not-the-formation logic that a serious adversary, and certainly China, has studied closely. Bazilian and Matisek are right that the domestic grid is a strategic vulnerability; I would add that in the Indo-Pacific theater specifically, it is very close to being the center of gravity.
The policy remedies the authors propose Defense Energy Security Zones with expedited FAST-41 permitting, aggressive use of the Defense Production Act to secure transformers and other long-lead grid components, and folding power resilience into Pentagon acquisition criteria are sound as far as they go, and Congress and the Department of War would do well to move on all three without further delay. But I would frame the ask differently, in a way that connects the grid conversation to the broader transformation the joint force is already trying to execute.
If JADO is genuinely the organizing concept for how the American military intends to fight, then the electrical grid that underwrites it has to be planned, funded, and defended with the same seriousness as any weapon system in the portfolio, and it has to be planned as a kill web asset, not a utility. That means treating grid resilience for critical defense nodes at home and, just as urgently, at Guam, in Australia, and across the first and second island chains as an explicit line item in the mesh force and impact force concepts the Marine Corps, the Navy, and allied planners are currently building. It means recognizing that intelligent mass cannot be surged into existence if the power behind the production line cannot be surged either, which argues for treating grid hardening and transformer stockpiling with the same urgency as munitions stockpiling. And it means acknowledging that an adversary practiced in chaos-inducing, infrastructure-first campaigns will not wait politely for Washington to finish a decade-long transmission permitting process before testing whether the American kill web actually has the resilience its architects assume it has.
There is also a maritime domain awareness dimension to this that the authors do not touch but that follows naturally from their argument. The Maritime ISR architecture I have been writing about, built around persistent unmanned surface and undersea sensing feeding a mesh of manned and unmanned shooters, generates exactly the kind of continuous, distributed, AI-processed data flow that Bazilian and Matisek identify as electrically ravenous. A MISR-centric force is, by design, trying to see more, more persistently, across a wider littoral space than a legacy, exquisite, few-large-platform Navy ever needed to. That is a feature, not a bug, of the mesh force concept but it is also a permanent increase in the electrical demand signal at every shore node, gateway, and processing center that supports it.
Firms building commercial maritime domain awareness capability on continuously ingested AI models are themselves quietly dependent on the same data center power constraints Bazilian and Matisek describe for the Pentagon’s own AI tools. The lesson generalizes: as maritime ISR becomes more autonomous and more continuous, the energy footprint of “seeing” becomes as strategically significant as the energy footprint of “shooting,” and both now sit downstream of the same balkanized, underbuilt grid.
None of this diminishes the value of the Bazilian-Matisek argument; it strengthens the case for taking it further than the authors themselves do. Their three recommendations, Defense Energy Security Zones, aggressive use of the DPA for grid components, and power resilience as an acquisition criterion, are necessary but, read against the kill web and mesh force logic, insufficient on their own.
A fourth recommendation follows naturally from the analysis above: allied grid resilience, particularly at the forward-deployed nodes that a mesh force and an impact force concept depend upon in the Indo-Pacific, needs to be planned and resourced jointly with allies such as Australia and Japan, not treated as a purely domestic U.S. infrastructure question. The Sir Richard Williams Foundation’s “Fight Tonight” work has repeatedly surfaced the same structural point in a different guise: readiness is not a snapshot of platforms on a ramp or hulls in the water, it is a standing capacity to regenerate combat power under contested conditions, and that capacity now runs through substations and transformers as surely as it runs through munitions depots and maintenance hangars.
Bazilian and Matisek have named the bottleneck correctly: the arsenal of the future runs on electrons before it runs on anything else.
The task now is to stop treating that insight as an energy policy footnote to military strategy and start treating it as what it actually is, the physical layer of the kill web, the precondition for intelligent mass, and, for the United States and its allies in the Indo-Pacific in particular, very possibly the single most consequential infrastructure vulnerability standing between current force design and the fight the joint and combined force says it is preparing for.
The Grid is the Arsenal: Power Wars and the New Foundations of Military Strength