The wires and buildings that carry modern computing are more exposed than their users usually realise, and the war has begun treating them as what planners call terrain. A data centre and a seabed cable are not ordinary industrial sites. They are single points whose failure spreads through services that have no local alternative: card payments, phone calls, cloud workloads, the coordination of everything from logistics to supply chain finance. The Ukraine war, and now the wider conflict, have shown both how attackable that layer is and how much resilience can be engineered into it before an attack arrives.

The cable network, by the numbers

The world's international data moves on a physical network of optical cables laid on the seabed. The system comprises several hundred cables in service today, stretching across roughly 1.4 million kilometres, funded by consortia of telecommunications companies, content platforms and investment funds. The finance has shifted over the last decade: the largest consumers of data, the cloud and streaming companies, now fund a large share of capacity because they need the routes for their own services.

Faults are routine and almost always accidental. Cable breaks happen many times a year, most often caused by anchors, fishing gear and seabed movement, and they are repaired by a specialist fleet of ships counted in the dozens. The system's resilience is not that cables never break; it is that traffic reroutes. A cable cut in the North Atlantic reroutes through the next cable within hours, with congestion rather than outage as the visible effect.

Two structural vulnerabilities complicate that comfort. The first is geography: cables converge on a small number of landing stations and shallow straits, and a coordinated action at a convergence point removes several independent paths at once. The second is repair capacity: in a scenario with multiple simultaneous faults, the fleet cannot be in several ocean basins at the same time, and delays are measured in weeks, which is why insurance for cable damage and government interest in repair ships have both grown sharply.

Data centres and the four dependencies

A data centre is a building wrapped around four systems: power, cooling, connectivity and the computing itself. The war's data centre episodes have demonstrated that the four can be attacked separately. Power is the most common target and the cheapest to disrupt, since a grid fault or a targeted strike on a substation reaches every machine in the building at once. Cooling is the most fragile engineering, because it depends on continuous water or air flow and fails in hours without it. Connectivity is the most substitutable, since routes can be shared across buildings. The computing hardware is the least attackable and the most replaceable.

Modern operators plan in tiers of redundancy, and the war has stress tested those plans in real time. Workload migration, power rationing and the physical dispersal of critical systems have kept services running in Ukraine through years of strikes. The same playbook, corporatised into resilience engineering and buy-in from insurers, is now visible in European and Gulf data centre projects: dual power feeds, on site generation, water independent cooling and geographic dispersal of the two facilities that must never fail together.

Why this layer is suddenly strategic

Three trends have promoted cables and data centres from technical assets to strategic ones. The concentration of compute in a small number of very large campuses means the failure of a site is a systemic event rather than a local annoyance. The growth of digital services means that a country's economy and its armed forces now run on the same networks, blurring the distinction between civilian and military infrastructure. And the technology to disrupt both layers has become cheap: a dragging anchor needs no missile, and a well-placed cyber intrusion needs no ship at all.

States have responded mostly by mapping and coordinating rather than by defending, because the assets are privately owned and span jurisdictions. NATO coordination on undersea infrastructure, national reviews of cable landings and the designation of critical operators are the visible instruments. The deeper change is in procurement: redundancy is being bought rather than assumed, and the repair fleet itself is being treated as strategic capacity, with new ships financed by governments and consortia that once regarded them as a cost.

The lesson the war teaches

The pattern across cable cuts, grid attacks and data centre strikes is consistent. Attacks on infrastructure rarely achieve the outage their planners imagine, because modern operators are good at redistributing load. They do, however, impose continuous cost: hardening, redundancy and insurance all show up as permanent expenses, and the fear of the next event shapes where new capacity is built. The strategic effect is not a day without internet; it is an internet that pays a war tax forever.

For governments the arithmetic creates an incentive to invest first in the layer that cannot be replaced: repair ships, landing station security, national power resilience and the connectivity corridors that no single actor can reach. For the rest of the economy, the useful realisation is more modest and more actionable: the four dependencies are known, they can each be tested, and the organisations that weathered the war's disruptions were the ones that had tested before anyone attacked.