The Internet Is on the Seabed

More than 99 per cent of intercontinental internet traffic travels along fibre optic cables lying on the seabed. There are several hundred of them, spanning well over a million kilometres.

Satellites are not an alternative. They matter enormously for coverage, for places cables do not reach, and for emergency restoration, and they carry a rounding error of the total by comparison. The cloud, for all the vocabulary built around it, is a few hundred cables on the bottom of the ocean, and the metaphor has been doing serious damage to public understanding for twenty years.

The vulnerability is not where people assume.

Cutting a cable is easy and always has been. What matters is repair, and repair is where the system is genuinely thin. The world's fleet of specialised cable ships numbers in the dozens, they are old, they are expensive, and they are busy. A fault is located, a ship is scheduled, it sails to the position, grapples the cable from the seabed, lifts it, splices it and lowers it back. This takes days in easy conditions and weeks in difficult ones, and if several faults occur at once the constraint is the queue for a vessel rather than the engineering.

Now the part that makes this a chapter in a geopolitics book rather than an engineering one.

Most cable faults are accidents. Anchors drag, trawls snag, undersea landslides happen. There are on the order of a hundred or more faults every year worldwide, and the overwhelming majority are mundane. That genuine background rate is precisely what makes deliberate damage attractive: a cut cable looks like the ordinary world, and attribution has to come from the movement of ships rather than from the damage itself. Ship tracking systems can be switched off. Anchors can be dragged unintentionally, and saying so is very hard to disprove.

Cable geography also explains a good deal of political geography, and a map of the routes is a better guide to who depends on whom than a map of alliances.

Landing stations cluster in a few favourable places, which gives those countries a structural position in global communications that survives every change of government. Routes designed to avoid a rival's waters are longer, costlier, and get built anyway, which tells you what the builders think the risk is. And the corridor through the Red Sea funnels a large share of Europe to Asia traffic through water that is narrow, shallow and contested, which is why damage there in early 2024 was felt on two continents at once.

The pattern from the chip chapter repeats exactly. Cables were laid where laying them was cheapest and routes were chosen for economics, one commercial decision at a time, by companies with no responsibility for the resilience of the whole. The concentration that resulted was nobody's plan and is now somebody's opportunity.