Intercontinental internet traffic runs almost entirely through fibre optic cables laid on the seabed. Satellites carry a small share, and the reasons are physical.
Fibre wins on capacity and delay
A single modern cable carries an enormous volume of data on many wavelengths of light in each fibre pair, which no radio link approaches.
Distance also matters. A signal to a high orbit and back adds a delay that is noticeable in a video call and severe for anything conversational.
Satellites remain essential for places cable cannot reach economically, and low-orbit systems reduce the delay, but the trunk routes between continents stay on the seabed.
The route is chosen for hazards, not distance
Cables avoid known fishing grounds, anchorages and unstable seabed where possible, so the laid path is longer than the straight line.
Near shore, where the water is shallow and traffic is heavy, the cable is buried in the seabed and armoured, because that is where damage happens.
Landing points are also constrained by permits and by the availability of a station on shore, which concentrates many cables into the same few beaches.
Breaks are routine and repairs are slow
Most faults come from fishing gear and ships' anchors rather than from anything dramatic, and they happen regularly somewhere in the world.
Repair means locating the fault electrically, sending a specialised ship, grappling the cable from the seabed, lifting both ends and splicing in a new section.
The ship may be days away and the weather may not cooperate, so a repair is measured in weeks, which is why capacity is bought across multiple cables.
Redundancy is a commercial arrangement
Networks survive a cut because traffic reroutes onto other cables automatically, provided capacity has been contracted on them in advance.
Regions served by only one or two cables have no such cushion, and a single break there produces slow connections and dropped services for the whole country.
The cost of diversity is why smaller markets are more exposed than the volume of their traffic alone would suggest.
Ownership has shifted
Cables were historically built by consortia of telecoms carriers that sold capacity to each other and to service providers.
Large content and cloud companies now fund cables directly, because they move enough of their own traffic to justify owning the route rather than renting it.
That shift changes who decides where new capacity goes, and routes increasingly follow the map of data centres rather than the map of national carriers.