More than 99 per cent of international data traffic descends from coastal landing stations into hair-thin glass fibres and crosses the ocean floor in darkness. TeleGeography now tracks more than 600 active and planned submarine cable systems, a number that changes as new routes enter service and older ones are retired.
A video call between Sydney and London may begin and end wirelessly, but almost everything between those two points travels through glass resting kilometres below the waves. Our earlier explainer on why intercontinental internet traffic travels by cable rather than satellite describes that hidden route.
The less visible part of the story is who now finances and uses it. TeleGeography reported that content and cloud networks accounted for 75 per cent of used international bandwidth in 2025, with networks operated by companies such as Google, Meta, Microsoft and Amazon among the largest sources of demand.
How a garden-hose cable carries continents
For most of its journey through deep water, a modern telecommunications cable is approximately as wide as a garden hose. The glass filaments carrying the signal are far thinner, roughly comparable to a human hair, and are surrounded by insulation, a power conductor and protective layers.
A cable usually contains several pairs of optical fibres. Lasers fire rapidly modulated pulses through those fibres, allowing newer systems to move hundreds of terabits per second between landing stations on opposite sides of an ocean.
The light cannot travel indefinitely without weakening. Submarine repeaters positioned along the route amplify it, commonly at intervals of around 100 kilometres, while electrical power is supplied from stations on shore through a conductor running inside the cable. Google’s Firmina cable was designed so that its repeaters could temporarily be powered from one end if the supply at the other end became unavailable.
Protection changes with depth. Near coasts, where anchors, fishing gear and currents present the greatest danger, cables are armoured and may be buried beneath the seabed. Across the abyssal plains, they are usually thinner and rest directly on the bottom.
Damage is routine rather than extraordinary. The International Cable Protection Committee records approximately 150 to 200 faults each year, with around 70 to 80 per cent caused by accidental human activity, primarily fishing equipment and ships’ anchors.
Why hundreds of cables can still be vulnerable
More than 600 systems sounds like an enormous safety net. It is not the same as having 600 completely independent routes.
Cables cluster along corridors chosen for seabed conditions, distance, permitting and access to terrestrial networks. The Luzon Strait, the Red Sea, the Mediterranean approaches to Europe and the Atlantic landing zones around Britain, France, Spain and the eastern United States carry unusually dense concentrations of infrastructure.
In early 2024, four named systems in the Red Sea lost service. Hong Kong operator HGC said the affected routes were SEACOM, TGN, Asia-Africa-Europe 1 and Europe India Gateway, estimating that the disruption touched about 25 per cent of traffic passing through the corridor.
The internet did not stop. Operators redirected traffic through other Red Sea systems, terrestrial routes across Asia and longer paths travelling east across the Pacific and Atlantic.
That rerouting is the system’s real redundancy. Large operators divide capacity among multiple cables, but the alternatives can be longer, more congested or exposed to the same regional hazard.
Landing sites narrow the map further. A cable needs suitable seabed geometry, permission to cross territorial waters, space for a landing station and a high-capacity terrestrial network waiting onshore. Once those facilities exist, later cables are naturally drawn towards them.
How technology companies became cable owners
For most of the twentieth century, international cables were financed by consortia of national and regional telecommunications carriers. Each participant paid part of the construction cost and received an agreed share of the finished capacity.
That arrangement suited an era when carriers sold international telephone calls and internet transit to other companies. Google, Meta, Microsoft and Amazon operate under a different equation because much of the traffic moving between continents belongs to their own services or links their own data centres.
At sufficient scale, repeatedly buying capacity from a carrier becomes less attractive than obtaining a fibre pair, committing to a large long-term capacity purchase or joining the group financing a new system. Ownership also allows a company to choose landing points and connect the route directly to its private backbone.
TeleGeography’s current inventory contains more than 60 cable investments associated with content providers. The arrangements range from sole ownership to consortium stakes, major capacity purchases and long-term rights to use individual fibres.
Google has the largest publicly documented portfolio. Its wholly owned systems include Curie, Dunant, Equiano, Firmina and Grace Hopper, alongside stakes in consortium cables and additional systems under construction.
Meta is one of the principal investors behind 2Africa, whose core infrastructure now reaches 33 countries. With its Pearls extension, the complete system is planned to span approximately 45,000 kilometres, while the company is also developing the even larger Project Waterworth.
Microsoft co-owns systems including Marea and Amitié and holds major capacity on others. Amazon has combined capacity purchases and consortium investments with dedicated projects, including the planned Fastnet route between the United States and Ireland.
What private investment changes
The immediate advantage is capacity. Technology companies have supplied billions of dollars for new fibre pairs, alternative landing points and routes that traditional carriers might not have financed at the same speed.
The same investment can improve resilience when a new cable follows a physically separate path. A second landing country or a route south of an established corridor gives operators somewhere else to send traffic during a fault.
The concern is not that one company can switch off the entire internet. It is that decisions about where new routes go, which data centres they connect and how quickly additional capacity is installed are increasingly made inside a small number of private infrastructure organisations.
Governments already treat those choices as national-security questions. In 2020, the US body commonly called Team Telecom recommended refusing the Hong Kong portion of the Pacific Light Cable Network, and the later licensed configuration connected the United States with Taiwan and the Philippines while excluding the proposed Hong Kong connection.
Ownership also requires careful wording. A company may own an entire cable, hold a minority consortium stake, own selected fibre pairs, lease capacity or make a large purchase before construction begins. Those positions provide influence and dedicated bandwidth, but they are not identical forms of control.
The cables themselves remain shared physical infrastructure with an expected design life commonly measured at around 25 years. During that period, operators must monitor faults, maintain spare cable and secure access to the specialised ships capable of raising a damaged section from the seabed.
The old geography beneath the new glass
The first transatlantic telegraph cable reached Ireland and Newfoundland in 1858. The Institution of Engineering and Technology’s archive records how the crews fought storms, snapped cable and repeated failures before the line finally worked, briefly linking the two continents. Make Tech Easier has explored how that three-week cable established the pattern followed by modern systems.
Today’s fibres are incomparably faster, but the broad geography remains recognisable. Routes still favour manageable seabed slopes, practical landing sites and the shortest safe connection between major communications markets.
When one fails, a repair vessel steams to coordinates calculated from electrical and optical tests. The crew grapples for a cable lying in darkness, lifts it through kilometres of water, cuts away the damaged section, splices in new glass and lowers the line back to the seabed.
The map will keep changing, but the physical act remains the same: light enters glass at one coast, crosses thousands of kilometres of black water and rises at another. The companies deciding where the next fibres land are now increasingly the same companies whose clouds, searches, feeds and software will fill them.
