The Cables On The Ocean Floor

AAsk most people how a message crosses an ocean and they will point at the sky. The answer lies on the seabed. Something close to ninety-nine per cent of intercontinental data traffic travels through fibre-optic cables laid on and under the ocean floor, satellites carrying the small remainder. The cables themselves are unimpressive objects. Across the deep ocean a modern one is about the width of a garden hose, its glass fibres wrapped in steel wire and polyethylene; only where it approaches a coastline is it thickened with armour and ploughed into the sediment. The first working telegraph line between Europe and North America was completed in 1866, and the industry has repaired breaks ever since without the public noticing. What has changed is the traffic. When a single pair of fibres can carry the working day of a mid-sized economy, the question of how easily such a line can be severed stops being a technical curiosity.

BNikolai Wickramasinghe, a marine geologist at the Nordkapp Institute of Ocean Engineering, has spent a decade arguing that the danger is banal. Between 2009 and 2021 her group assembled a register of 1,340 faults reported by operators in the Atlantic and the Indian Ocean, each matched against water depth, seabed type and, where a repair ship recovered physical evidence, the object responsible. Roughly seventy-one per cent of the faults occurred in water shallower than two hundred metres, a band that accounts for a small fraction of the total length of cable in service. Fishing gear dragged along the bottom and ships' anchors together explained most of them. Wickramasinghe's conclusion is not that the sea is hostile but that the traffic above it is: outside the shelf, she maintains, a cable is left alone for decades. Thickening the cable, on her account, treats the symptom; moving the route, or burying it deeper, treats the cause.

CThe argument makes a prediction, and Wickramasinghe has tested it. If human activity is the dominant hazard, then faults in deep water should be rare, and the rare ones should have natural causes that leave a geological signature. Her survey of the deep-water record found 38 faults over the same period, of which 29 were traced to turbidity currents, avalanches of sediment that run down submarine canyons at speed and can cut several cables in sequence, as one did off the Cape Verde rise in 1961. Damage of this kind, she notes, is confined to margins where a large river discharges sediment directly into the head of a canyon. Elsewhere in the deep the register is close to empty. The asymmetry matters for planning, because a hazard concentrated in known corridors can be designed around, whereas one spread evenly across the seabed cannot. Wickramasinghe presents the finding as a constraint on where cables should be laid rather than as a reassurance about the system as a whole.

DMaja Zielinski, an economist of network infrastructure at the Accra Centre for Digital Systems, accepts the fault statistics and rejects what is made of them. Counting breaks, he argues, measures the sea; what matters to a country is the interval between a break and its repair, and that is decided on land. His study of nine island and coastal economies put the median wait for a repair vessel at 19 days, of which only four were spent at sea; the rest went on securing permits to work in national waters, a delay that has lengthened since 2015 even as the fleet has shrunk. Twenty-one ships worldwide are equipped for deep repair, and their mean age is twenty-seven years. Zielinski concedes two weaknesses. His nine cases were chosen because their outage records were public, which may over-represent badly served places, and the cost figures came from operators themselves, since no independent audit of the sector exists.

EThe disagreement is no longer confined to journals. Several governments introduced licensing rules in 2023 requiring new cables to make landfall at more than one station, and insurers have begun to price routes by their exposure to narrow passages where dozens of systems share a corridor a few kilometres wide. Zielinski points out, awkwardly for his own case, that repair capacity cannot be legislated into existence quickly, since a cable ship takes about three years to build and its crews longer to train. Wickramasinghe replies that the two questions cannot be separated in practice, because a regulator reading a low fault count concludes that the system is sound, while one reading a long outage concludes that the seabed is dangerous. Both regard the habit of reporting individual dramatic failures, rather than the ordinary annual record, as having done the subject harm.

FA settlement of sorts is taking shape in which the two accounts are assigned to different stretches of the same cable rather than ranked against one another. On the continental shelf the hazard is human, predictable and largely met by burial and by routing around fishing grounds; in the deep ocean breaks are rare enough that the cost of an outage is set almost entirely by how long a ship takes to arrive. The unresolved question is whether the redundancy that exists on paper survives contact with geography, since two cables described as independent may share a landing station, a canyon or a strait, and no public register now records those overlaps. What both researchers reject is the assumption behind a decade of reassuring headlines, namely that a network with many links must be a network with many routes.