If the Cable Breaks, Can Satellites Take Over?

How does international internet traffic actually travel?

Overwhelmingly through submarine cables. This surprises many people, because the internet they use every day feels wireless, but the wireless part covers only the last short hop. The ocean crossing is almost entirely physical fibre.

There are fewer cables than people assume and far fewer landing points. Bringing a cable ashore needs suitable seabed, a protected stretch of coast, and a local landing permit, so several cables serving one region often converge on a handful of landing stations. That is a textbook single point of failure: many routes on paper, a few doorways in reality.

Satellites operate at an entirely different scale. Even the large low-earth-orbit constellations of recent years have a total capacity only on the order of one or two new submarine cables, and that capacity is shared among users across the whole globe; what reaches any one region is far less than one cable carries.

How do cables break, and how long do repairs take?

Most breaks are accidental: fishing and ships’ anchors account for roughly 70–80%, followed by seabed earthquakes and landslides. Some 150 to 200 faults occur worldwide each year, and most go unnoticed because traffic reroutes automatically.

Repair requires a specialised cable ship. There are not many, they are unevenly distributed, and the work involves locating the fault, lifting the cable off the seabed, splicing it, and laying it back. That takes weeks when it goes well, and longer when weather, sea conditions, or permits intervene. This is the sharpest contrast with networks on land, where a broken line can be dug up the same day.

It is also why deliberate damage is hard to tell from accidental damage in the moment. The same marks can come from a trawler or from an anchor dragged on purpose, and attribution takes time and evidence. Grey-zone operations favour this method for that reason: the damage is real, responsibility is blurred, and the victim has to prove its case before it can respond.

What did low-earth-orbit constellations change?

Two things above all: latency and speed of deployment. Low orbits are far closer to the ground, cutting round-trip latency from hundreds of milliseconds in geostationary orbit to tens, which makes real-time calls and remote operation practical. A terminal needs only an antenna to get connected, with no physical line at all.

For the first time, satellites became a genuine backup rather than merely an option for remote areas. That is what makes them valuable in a crisis: they depend on no local physical infrastructure, so they do not fail along with it.

The capacity gap did not change, though; it only became less obvious. Constellations suit traffic that is small in volume and high in importance, such as command communications, financial settlement, emergency coordination, and news transmission. They are not suited to carrying an economy’s everyday internet. Treating them as an equivalent substitute for cable is the most common over-optimism on this subject.

Constellations are also a textbook dual-use system: a single constellation serves commercial users and military units at once. That complicates their position in a conflict, because interrupting a commercial service and disabling a military capability become technically the same act.

Who decides when the service stops?

Beyond capacity, this is the most fundamental difference. Submarine cables are often owned by consortia of carriers, which makes it hard for any one of them to shut down service. A constellation typically sits with a single operator, which can enable, restrict, or geographically limit service from its own console.

So resilience takes a different shape in each. A cable’s weakness is physical: an anchor can sever it, and repair is slow. A constellation’s weakness lies in governance: nothing severs it, but service in a region can end through a commercial decision, an export control, or a policy of the operator’s home state.

“We have a backup” therefore needs qualifying. Physical redundancy and governance redundancy do not substitute for one another: another cable does not solve a single-operator problem, and another satellite contract does not solve a seabed one. In this context, resilience means having both, and confirming in advance that they do not share a weakness.

The logic is exactly that of a chokepoint, except that the choke points are landing stations and administrative consoles rather than straits: whoever holds such a point holds asymmetric bargaining power.

What does this mean for Taiwan?

Taiwan is an island whose international connectivity rests almost entirely on submarine cables, which come ashore at a small number of points. The single point of failure described above is therefore concrete rather than abstract. The outlying islands are more exposed still: when their cables are all cut, internet access can become badly congested, as has already happened.

Preparation works on two layers. The physical layer means more landing points and more diverse routes, including deliberately indirect paths in different directions. The governance layer means making sure backup is not concentrated in one operator and agreeing crisis-time service terms in advance; leave the negotiation until the crisis, and the outcome depends on how quickly it can be concluded.

For an ordinary organisation, two questions are worth keeping in mind: can critical operations continue with connectivity fully down, and is the “backup” merely a second account on the same path? The second is the same question the de-risking explainer asks: diversification on paper and diversification in the physical chain are often not the same thing.