Nearly all intercontinental internet traffic travels through cables lying on the seabed. When one breaks, the repair is a maritime operation that takes weeks rather than hours.
Most breaks are caused by human activity
Fishing gear dragged across the seabed and ship anchors account for the majority of faults, which is why breaks cluster in shallow coastal waters and busy approaches.
Cables in those zones are armoured and often buried, but burial depth is limited by seabed conditions and an anchor dropped in an emergency can reach far below it.
Deep-water faults are rarer and usually caused by seabed movement or equipment failure, but they are considerably harder and slower to reach for repair.
The fault is located electrically before any ship sails
Operators send a test signal from a terminal station and measure the reflection returned from the discontinuity, which indicates the distance along the cable to the fault.
That distance is converted to a seabed position using the cable's as-laid route, which is surveyed and recorded precisely at installation for exactly this purpose.
Accuracy matters because the repair ship must find a specific cable on a featureless seabed, and searching an imprecise area consumes days of expensive vessel time.
Repair requires a specialised vessel
A small global fleet of cable ships carries splicing equipment, spare cable and remotely operated vehicles, and vessels are positioned under shared maintenance agreements covering defined regions.
The ship grapples the cable, cuts it, and raises each end to the deck in turn, buoying one end while the other is recovered.
The fibres are then spliced and tested, spare cable is inserted to replace the damaged section, and the repaired length is lowered back with a deliberate slack loop.
Weather and permissions dominate the timeline
Splicing requires workable sea conditions, so a repair scheduled in a stormy season can wait at anchor for a suitable window.
Work inside a country's waters generally requires permits, and the time to obtain them varies enormously between jurisdictions, sometimes exceeding the technical work itself.
These factors, rather than the splice, explain why a break announced one week may not be restored until several weeks later.
Redundancy hides most faults from users
Traffic is routed automatically over surviving paths when a cable fails, so users on well-connected routes may notice nothing beyond marginally higher latency.
Places served by only one or two cables have no such cushion, and a single break there can reduce national connectivity substantially until repair.
This asymmetry drives investment in additional landing points, since the resilience of a connection depends on the number of independent paths rather than the capacity of any one of them.