The shortest route is under water
A message crossing an ocean is unlikely to travel by satellite. It usually becomes pulses of light moving through glass fibres on the seabed. Submarine cables carry more than 99 per cent of international internet traffic, according to the International Telecommunication Union.¹ The apparent weightlessness of the internet depends on long, vulnerable objects that ships, fishers and repair crews can actually touch.
A cable begins with a surveyed route between landing points. Specialist ships pay it out across the seabed. Near coasts, where anchors and fishing gear are most likely to reach it, the cable may have extra steel armour and be buried beneath sediment. In deep water it may rest on the seafloor. At its centre are hair-thin optical fibres; on long routes, powered repeaters amplify the fading light signal. The outer layers protect the fibres and carry the electrical power needed by that equipment.²
This engineering is a compromise. A cable must survive pressure, corrosion and handling, but also be flexible enough to lay across complex terrain and later lift for repair. Route planners try to avoid hazards such as steep slopes and busy anchorages. They cannot avoid the entire working ocean.
A break is a physical and a network event
Many faults come from ordinary maritime activity. Fishing equipment can snag a cable; a dragging anchor can crush or pull it. Earthquakes and undersea landslides can damage several nearby systems at once. The International Cable Protection Committee describes anchor damage as a significant, preventable cause of faults, while the ITU identifies fishing, anchoring and natural hazards among the principal risks.³⁴ A break need not be a clean severing: bent glass, damaged insulation or a failed repeater can also interrupt service.
Operators first detect the fault through changes in signal or power at the cable's landing stations. Tests from shore help estimate its distance along the route. The location is therefore not found by a diver searching an ocean-sized haystack. A repair vessel must still sail to the area, obtain permits where required and work around weather and sea conditions. TeleGeography's account of the process describes grapnels used to lift the cable and technicians splicing replacement fibres aboard a specialised ship.⁵
A damaged section is commonly brought to the deck, cut away and replaced with spare cable. Each tiny glass fibre must be joined accurately; the cable is tested before being lowered again. In shallow areas, a remotely operated vehicle may rebury it.⁵ Repair is closer to precision surgery at sea than to plugging in a spare lead. The time taken varies with vessel availability, depth, weather and permission to work in a particular jurisdiction.
Does the internet go dark?
Often, users notice nothing. Cable owners and network operators buy capacity across multiple routes, so traffic can move to another cable when one fails. Routing between networks can also change to reach the same destination by a different path. A documented Jersey outage showed traffic shifting from a direct London path to a backup route via Paris under the internet's inter-network routing system.⁶
But rerouting is not a magic spare ocean. Alternate paths have finite capacity and may be longer. Speeds can fall, latency can rise, and some services may be unavailable while operators rebalance traffic. If several cables share the same narrow corridor or landing area, one incident can disable more than one escape route. Cloudflare documented how cuts to two cable systems in 2022 disrupted users and services across parts of Africa, the Middle East and Asia.⁷ The consequences depend on the actual topology: how many paths a region has, where they land and who can use them.
Small islands and less connected countries can be especially exposed. A place with only one or two practical international links has fewer ways around a fault than a large hub with many landing stations. The ITU's 2026 resilience work highlights geographic concentration and reliance on a small number of cable systems as major weaknesses.⁸ Satellite links can help in particular situations, but they do not simply replace the capacity of the global fibre network.
A strategic map nobody sees at the beach
The routes reflect economics as well as geography. Cables are financed and used by telecoms operators, cloud companies and other investors; landing rights and repair permissions involve governments. A cable that looks like one line on a map is part of a wider system of landing stations, data centres and commercial agreements. Protecting that system has become a matter of national resilience, because banking, healthcare, government services and daily communication depend on it.¹
Strategic concern should not turn every failure into a mystery. Accidents account for many breaks, and the cause of an individual fault may be uncertain until investigated. Sensible protection includes clear charting, safer anchoring, route diversity, monitoring and repair arrangements that work across borders. The ITU and cable protection industry have called for quicker repair processes and less geographic concentration.⁸
The most impressive thing about an undersea cable break is often its invisibility. The ship and splicing crew do painstaking work while packets quietly take another route. When that alternative does not exist, the ocean's hidden wiring becomes suddenly visible in dropped calls, slow payments and stalled services. The internet is resilient because people have built multiple physical paths; it is fragile wherever they have not. Building that resilience means paying for spare routes before a failure makes their value obvious.