Actuarial Warfare and the Geopolitics of Compute

Today's article follows up on my last Friday's post, where I introduced the concept of actuarial warfare as a new tactic in modern warfare. The idea is simple: launch multiple small attacks against critical assets of your enemy, such as oil tankers or data centers, not with the goal of destroying them but of making them uninsurable. Moreover, just publishing a list of assets prone to be attacked makes them uninsurable, which can strand the assets.

Where We Come From

For much of the past two decades, the technology industry operated under a comforting assumption: that the digital economy existed largely outside the traditional dynamics of geopolitics. The cloud was presented as a borderless infrastructure layer, abstracted from territory and politics, capable of distributing computing resources across the globe with unprecedented resilience. In this vision, digital infrastructure was neutral. Data moved across networks, applications ran on distributed servers, and the underlying architecture of the internet made geography increasingly irrelevant.

Recent events suggest that assumption is no longer sustainable.

When Iranian drones reportedly struck Amazon Web Services facilities in the Gulf on March 1st, the operational disruptions were relatively limited and quickly mitigated. Cloud architectures are designed with redundancy precisely to absorb localized failures. What matters more than the immediate damage is the precedent these attacks set. For the first time, hyperscale cloud infrastructure appears to have been treated as a strategic target within an interstate conflict. The episode revealed something that had long been implicit but rarely acknowledged: the infrastructure of the digital economy has become deeply entangled with geopolitical power.

This shift signals the emergence of what might be described as actuarial warfare — a form of strategic competition in which adversaries target infrastructure whose destruction produces economic or systemic consequences vastly disproportionate to the cost of attacking it.

Few infrastructures illustrate this asymmetry more clearly than modern data centers.

The Physical Reality of the Cloud

The metaphor of the cloud has always obscured an important physical reality. While software and digital services may operate across distributed networks, the infrastructure that supports them remains firmly anchored in the physical world.

Data centers are large industrial installations. They require vast amounts of electricity, high-capacity fiber connectivity, advanced cooling systems, land access, and complex permitting processes. Their locations are often visible through satellite imagery, and their power substations and network interconnections are readily identifiable.

In other words, the cloud is not an ethereal digital layer floating above geopolitics. It is a collection of buildings connected to power grids and telecommunications networks. Those buildings exist within sovereign territories, and those territories exist within geopolitical systems.

Once infrastructure reaches a certain level of economic importance, it inevitably becomes part of geopolitical competition.

The Economics of Asymmetric Disruption

Hyperscale data centers represent some of the most capital-intensive infrastructure projects in the modern economy. A single campus can require investments ranging from one to ten billion dollars and take several years to design, permit, and construct. These facilities depend on complex global supply chains that include high-voltage electrical equipment, cooling technologies, and increasingly scarce advanced semiconductors.

Yet the cost of disrupting such facilities can be dramatically lower. Targeting the electrical infrastructure that powers a data center, damaging cooling systems, or severing critical fiber links may temporarily disable systems whose construction required years of investment. Even if the facilities themselves remain structurally intact, interruptions to power or connectivity can cascade across the digital services that depend on them.

This imbalance between the cost of building infrastructure and the cost of disrupting it is precisely what characterizes actuarial warfare. In actuarial terms, the expected losses generated by a successful disruption can be far greater than the resources required to carry it out.

Centralization and Strategic Vulnerability

The design of the modern cloud has intensified this vulnerability in subtle ways. Over the past decade, computing capacity has become increasingly concentrated in large hyperscale facilities operated by a relatively small number of global providers.

This centralization has delivered enormous efficiencies. Hyperscale operators benefit from economies of scale, optimized energy usage, and the ability to deploy massive clusters of GPUs required for training and running artificial intelligence models. Concentrating infrastructure also simplifies operations and allows providers to build specialized facilities designed specifically for high-density computing.

However, this concentration also creates strategic points of failure. When billions of dollars of computing infrastructure and critical digital workloads are located within a handful of facilities, those facilities become high-value targets. Disruptions to a single cloud region may propagate across multiple industries, affecting financial systems, logistics networks, government services, and digital platforms simultaneously.

The cloud architecture was designed primarily to withstand technical failures such as hardware outages, software bugs, or localized natural disasters. It was not originally designed to withstand coordinated geopolitical targeting.

Digital Sovereignty and Infrastructure Control

The growing strategic importance of digital infrastructure also reshapes the discussion around digital sovereignty. Much of the debate over the past decade has focused on issues such as data localization, privacy regulation, or software ecosystems. Yet sovereignty in the age of artificial intelligence increasingly depends on a more fundamental capability: control over infrastructure.

A nation whose critical digital services depend on infrastructure that can be disrupted — whether by sanctions, cyber operations, or physical attacks — does not fully control its digital destiny. Infrastructure ownership, operational control, and geographic resilience therefore become central components of sovereignty.

As artificial intelligence systems grow more central to economic competitiveness and national security, the infrastructure required to support them becomes correspondingly strategic. Data centers, fiber networks, semiconductor supply chains, and electrical grids collectively form the backbone of what might be described as the geopolitics of compute.

Signals from Insurance Markets

Insurance markets often provide an early indication of emerging systemic risks. For years, insurers were hesitant to provide coverage for large-scale digital infrastructure risks because the potential losses were difficult to quantify. Cloud outages can cascade across sectors, creating complex chains of financial exposure.

Recently, however, insurance facilities specifically designed to underwrite AI infrastructure risk have begun to appear. These initiatives suggest that financial markets increasingly recognize the scale of the exposure associated with hyperscale infrastructure.

Yet most policies still contain standard war exclusion clauses, which exclude coverage for damage resulting from acts of war or state-sponsored attacks. As a result, the largest risks facing cloud infrastructure remain effectively uninsurable.

This reality implies that the economic burden of geopolitical risk will ultimately fall on operators, investors, and governments themselves.

The Future Geography of Compute

If geopolitical risk becomes a permanent feature of the digital infrastructure landscape, it may eventually reshape where and how data centers are built.

Regions that once appeared attractive because of favorable regulation, cheap electricity, or strategic connectivity may face new scrutiny if they are located near potential conflict zones. Conversely, inland regions with greater geographic depth and political stability may become more attractive for long-term infrastructure investments.

In addition, some operators are exploring hardened architectural approaches, including underground facilities or enhanced physical protection systems. Others are considering more distributed computing architectures that reduce reliance on a small number of massive hyperscale installations.

Interestingly, the original architecture of the internet was designed with this principle in mind: to avoid single points of failure by distributing network functionality across many nodes. The next phase of AI infrastructure development may rediscover some of these early design principles.

The End of the Peacetime Cloud

For most of its history, the cloud industry expanded during a period of relative geopolitical stability. Infrastructure planning focused primarily on technical resilience, operational efficiency, and energy costs. The possibility that hyperscale infrastructure might become a direct target in geopolitical conflict received comparatively little attention.

That environment appears to be changing.

Artificial intelligence is rapidly becoming a central driver of economic power and geopolitical influence. As that transformation unfolds, the infrastructure that supports AI — data centers, chips, power systems, and networks — will inevitably attract greater strategic attention.

In that sense, the digital economy is entering a new phase in which compute infrastructure begins to resemble energy infrastructure in its geopolitical significance.

The cloud industry built its architecture for an era of globalized stability. The emerging geopolitics of compute suggests that the next era may require a different kind of resilience — one designed not only for technical failures, but also for the strategic pressures of an increasingly contested world.