Picture yourself in a network operations centre somewhere in Frankfurt at two in the morning. Latency graphs that should be flat are spiking in long, ugly curves. Your colleagues in Mumbai are seeing the same thing. Nobody has touched a configuration file. Nobody needs to: something on the seafloor near Alexandria has been dragged, cut, or buried, and a corridor that carries roughly a third of all data moving between Europe and Asia is now doing the work of a fraction of itself.
The question that keeps network engineers quietly anxious isn't whether the cables are redundant. They are, technically. The question is whether the redundancy actually helps when the chokepoint itself is the problem.
Why the geography won't bend to the engineering
Underwater cables don't follow straight lines. They follow the seabed: avoiding trenches too deep to repair, skirting volcanic ridges, threading through shallow continental shelves where a ship's anchor can reach them. The result is that intercontinental routes funnel through a handful of natural passages, and those passages are dictated by plate tectonics, not network architects.
The Red Sea and the Suez approach is the clearest example. To get a cable from a landing station in southern Europe to one in India or Southeast Asia, you have essentially three options: through the Red Sea and the Bab el-Mandeb strait, around the Cape of Good Hope (adding roughly 20,000 kilometres and unacceptable latency for time-sensitive traffic), or over land through Central Asia (expensive, politically complicated, and still not widely built out at scale). So cables pile into the same narrow corridor. A dozen or more major systems share roughly the same stretch of shallow, busy, heavily fished water near Alexandria and Suez. When a cable ship drags an anchor across that patch of seafloor, it doesn't cut one link. It cuts several simultaneously.
The same logic applies, with local variations, to the Luzon Strait between Taiwan and the Philippines, the Strait of Malacca, and the approaches to the English Channel. Each one is a place where ocean physics and route economics conspire to produce a single corridor that everyone uses.
This is where the redundancy argument gets uncomfortable.
Operators do build redundancy. There are now more than 400 active submarine cable systems worldwide, and major trans-Atlantic routes carry traffic across eight or ten distinct cable paths. But redundancy across a route is not the same as redundancy across a chokepoint. If ten cables all pass through the same fifty-kilometre patch of seafloor, they are ten cables with one vulnerability. Think of it as ten different motorways that all, for reasons of geography, merge into a single tunnel for the last mile: the traffic engineers are not wrong that the roads are separate. They are wrong to call the system resilient. A single incident near Alexandria once cut three cables in four days and disrupted internet service across India, Egypt, and parts of the Gulf. The cables were technically separate systems. The failure was singular.
Consider two engineers, both working for telecoms companies that spent heavily on diverse routing. Call them Priya and Markus. Priya's traffic runs on a cable that lands in Marseille and threads through the Red Sea. Markus's traffic runs on a different cable, a newer one, built five years later with a different investor consortium. It also lands in Marseille. It also threads through the Red Sea. Their budgets were spent on redundancy. Their risk was not reduced. That is not bad luck. It is the predictable consequence of an incentive structure that rewards getting a cable built over building it differently.
The economics that keep the trap set
Building a submarine cable costs somewhere between $100 million and $500 million depending on length and depth, and those figures don't include landing rights, regulatory approvals in every country whose territorial waters you cross, or maintenance contracts. Route decisions are therefore intensely conservative. You go where the permissions exist, where repair ships can reach, where the landing stations are already built. And the landing stations are already built at the chokepoints, because that's where previous generations of cable builders went for the same reasons.
The infrastructure justifies the route. The route justifies the infrastructure. The chokepoint deepens with every iteration. It is a trap, and the door is made of sunk costs.
Satellites, including low-Earth-orbit constellations, are sometimes offered as the answer. They aren't. Not yet, not for bulk intercontinental traffic. Bandwidth per satellite link remains orders of magnitude below a modern fibre cable, and latency, while better than geostationary, is still higher than a direct undersea path. Anyone telling you otherwise is probably selling you a press release.
So ask yourself: if the engineers know where the vulnerabilities are, and the operators know, and the governments increasingly know, why does the map keep looking the same? Because knowing where the trap is and being willing to pay to escape it are two entirely different problems.
What people get wrong is assuming that more cables automatically means more resilience. It can. But only if the cables are genuinely geographically diverse, which requires someone to pay the premium for a longer, harder, more politically fraught route. That premium is real. The incentive to pay it is abstract.
The internet is often described as a system designed to route around damage. That's true of the logical network. The physical one is a different matter: a set of glass threads on a seafloor shaped by forces that predate the concept of redundancy by about 200 million years. The gap between those two descriptions is where the actual risk lives.