Submarine cables · Luzon, 2006 · Red Sea, 2024

II

The part that usually surprises is the first sentence of any industry report, and it is not a metaphor: more than ninety-five per cent of data traffic between continents goes by submarine cable. Satellite moves the rest.

There are somewhat more than five hundred active cables and around a million and a half kilometres of laid line. An intercontinental cable in deep water is the thickness of a garden hose; inside it carries a bundle of fibres the thickness of a hair, surrounded by steel and by a copper tube that conducts the current feeding the optical repeaters every sixty or hundred kilometres.

In the flat, deep zones it is left resting on the bottom, unburied. Burying it costs money and is only done near the coast, where there are anchors and trawl nets.

With that in mind, the statistics of faults stop looking exotic. Between one hundred and two hundred cuts a year are recorded worldwide. Most are caused by an anchor or a fishing rig; then come movements of the bottom, and lastly equipment failures. Almost nothing is sabotage.

They are not rare accidents. They are the normal operating regime.

III

What turns an ordinary fault into a regional blackout is geography, because the cables are not spread across the sea; they are bunched where the sea lets them through.

The Luzon Strait —between Taiwan and the Philippines— is where almost all the traffic of northeast Asia comes in and goes out. The Strait of Malacca concentrates the route between the Indian Ocean and the China Sea. The Red Sea corridor, from Bab el-Mandeb to Suez, funnels the traffic between Europe and Asia, and besides takes it out of the water: in Egypt the cables cross the country in a duct, which is the only stretch of the whole journey that does not go by sea. Those same cables narrow again afterwards in the Sicily Channel — two necks in series on a single route, which do not compensate for one another; they add up.

They are the same passages that have been strategic for ships for centuries, for the same reason: it is where you can go. A cable between Europe and Asia that does not want to go round Africa has to pass through the Red Sea, and in the Red Sea there is a stretch where fifteen or sixteen cables run almost together.

Baran had calculated that a network with three or four links per node survives the loss of half its nodes. The calculation is correct, and it assumes the links are independent. Sixteen cables that share two hundred kilometres of sea bed are not sixteen links: they are one.

IV

In February 2024 three of them were cut in the Red Sea — Seacom, AAE-1 and EIG — within the space of a few days.

The most accepted explanation is that the Rubymar, a cargo ship hit by a Houthi missile on 18 February, let go its anchor on being abandoned and dragged it along the bottom while adrift, before sinking. There is no definitive public confirmation, and it is as well to put it that way: it is the hypothesis the operators hold, not an established fact.

The effect has been measured. Around a quarter of the traffic between Europe and Asia had to be diverted, in part round the other side of the planet, with the increase in delay that implies.

The repair was not quick, and not for lack of technique. Permission from the Yemeni authorities was needed to work in those waters, and a ship willing to enter an area where ships were being shot at was needed.

V

There is a last stretch almost nobody sees, and it is where the cable comes out of the water.

A landing station is a low building, with no sign, a few hundred metres from any beach. Inside are the units that feed the cable with the direct current that drives the repeaters, and the point where the fibres from the sea are spliced to those on land. There is usually a fence, a camera and nobody inside most of the time; it is visited for maintenance, and little else.

In Spain they are at Conil, at Sopelana, at Altavista. In the United Kingdom, at Bude, next to a government signals listening station — it is not a coincidence, it is the same geographical reason applied twice.

And there is a change of ownership worth noting, because it moves the whole drawing. The cables used to be laid by consortia of telecommunications operators who shared out the capacity. Since the middle of the twenty-tens, those who pay for the new ones are in good part four content companies [the same ones that have the clouds], which no longer rent other people's fibre and lay their own.

The sea floor also concentrated, and it did so into fewer hands than before.

VI

The fleet remains, which is the part almost never talked about and the one that decides how long all of the above lasts.

The ships capable of repairing a submarine cable in deep water number some dozens in the whole world; the industry's counts come to around sixty, and a proportion of them are occupied laying new cable rather than repairing it. Many are over twenty years old. Hardly any are built, because the business consists of charging an availability fee and waiting, and that does not justify a shipyard.

The procedure is that of 1870, improved. The ship goes to the estimated position, drags a grapnel along the bottom until it hooks the cable, raises it, cuts, splices in a new stretch and returns it to the sea. At four thousand metres' depth, each pass is hours.

When eight cables break at once, eight ships do not go out. The one there is goes out, and does them one after another.

The contract that holds all this up is a maintenance agreement by zone: the operators of a region pay an annual fee to a consortium that keeps ships stationed and available. It works reasonably well when the faults are the usual ones, spread out in time. It is not sized for a whole strait breaking.

There are fewer ships in the world capable of repairing a submarine cable than passenger aircraft taking off from a large airport in an hour.