Chapter 7: The Cable War
Three hundred meters below the surface of the Red Sea, in the western channel of the Bab el-Mandeb Strait, a cable the diameter of a garden hose lies on the ocean floor. It is sheathed in polyethylene, armored with galvanized steel wire, and inside...
The Cable War
Three hundred meters below the surface of the Red Sea, in the western channel of the Bab el-Mandeb Strait, a cable the diameter of a garden hose lies on the ocean floor. It is sheathed in polyethylene, armored with galvanized steel wire, and inside that armor, surrounded by a copper conductor that powers the repeaters every seventy kilometers, are hair-thin strands of glass that carry pulses of light at close to the speed of light itself. Each pulse represents data — a financial transaction, a medical record, a video call, an AI inference query, a training dataset in transit, the ordinary and extraordinary communications of a connected world. This single cable carries more information per second than the entire global internet did in the year 2000.
The water here is dark and cold. The pressure is thirty atmospheres. The cable was designed to last twenty-five years on the seabed without maintenance. Nothing about its engineering anticipated what is happening on the surface above it: a war in which the combatants have discovered that the cheapest way to disrupt the infrastructure of the digital age is to sink a ship and let the anchor drag.
Fifteen submarine cables carrying seventeen percent of global internet traffic — over ninety percent of all Europe-Asia data capacity — pass through the Bab el-Mandeb Strait. The strait is twenty-six kilometers wide at its narrowest. Its name, in Arabic, translates to "Gate of Grief." Telecom consortiums have invested billions of dollars to lay these cables. They all converge in the same chokepoint because geography offers no alternative: there is no submarine path between the Indian Ocean and the Mediterranean that does not pass through the Red Sea, through the Bab el-Mandeb, and through Egyptian territory on the far end. The only bypasses are to circumnavigate Africa — adding enormous distance and latency — or to wait for Arctic cables that will not be operational until 2027 at the earliest.
Every cable. Every bit of data. Through one gate.
On February 18, 2024, Houthi ballistic missiles struck the MV Rubymar, a Belize-flagged, British-owned cargo ship carrying 41,000 tons of fertilizer, off Yemen's west coast. The crew abandoned ship and dropped the anchor. The anchor failed to hold. For six days, the Rubymar drifted, dragging its anchor across the seabed.
At 09:46 UTC on February 24, the SEACOM cable was severed. Five minutes later, at 09:51, the AAE-1 cable was severed. The EIG cable was cut in the same window. Three cables, worth a combined $3.5 billion, destroyed by a piece of steel dragging across the ocean floor behind a sinking ship.
The impact was immediate. Twenty-five percent of data traffic between Asia and Europe disrupted. Seventy percent of Europe-Asia data flow affected on some routes. Over one hundred million people in West and North Africa experienced degraded internet service. Internet services in Djibouti went down entirely. Microsoft Azure confirmed increased latency for Middle East cloud services. "Like plumbing," said Doug Madory, a network analyst. "You lose some volume of water coming down the pipes."
The Houthis denied deliberately targeting the cables. Their leader, Abdulmalik Al-Houthi, said so on television. The evidence supports the anchor-drag theory — GPS data of the Rubymar's drift, the five-minute gap between cable cuts consistent with an anchor scraping across the seabed, the well-established fact that anchor drag is the number one cause of cable damage globally. This was not, by the available evidence, a deliberate attack on undersea infrastructure.
It did not need to be. The Houthis attacked the ship. The ship dragged its anchor. The anchor cut the cables. The distinction between deliberate infrastructure targeting and incidental collateral damage from anti-shipping operations is strategically irrelevant. You do not need to aim at the cables if your missiles create the conditions for cable damage every time they find a target.
The repair took months. Yemen's split government — the internationally recognized administration in Aden, the Houthi-controlled authority in Sana'a — created a permitting nightmare worthy of Kafka. Cable repair ships needed authorization from both factions. The Yemeni Ministry of Information and Technology required licenses from the Sana'a Maritime Affairs Authority, which is controlled by the group responsible for the attacks. Months of negotiations followed. The E-Marine vessel Niwa was deployed. AAE-1 was repaired in July 2024, roughly five months after the cut. Full capacity to Hong Kong and Singapore was not restored until December 31 — ten months.
In March 2025, the PEACE cable — fully Chinese-built and owned by the Hengtong Group, connecting Chinese assets in Gwadar, Pakistan, to Djibouti, which hosts a Chinese naval base — was cut 1,450 kilometers from the Egyptian coast. It was repaired in three weeks by HMN Tech, formerly Huawei Marine. Chinese infrastructure, maintained by Chinese capabilities, restored on a Chinese timeline.
In September 2025, four more cables were severed near Jeddah, Saudi Arabia — SMW4, IMEWE, FALCON, and EIG — attributed to "commercial shipping activity," likely another anchor drag. Microsoft Azure again confirmed latency surges. Internet degradation across India, Pakistan, and the UAE. Users on Du and Etisalat, the two largest Emirati telecoms, reported slower speeds and intermittent access.
Three incidents in eighteen months. Seven cables damaged. And the system held. This is worth saying clearly: the internet did not go dark. The redundancy within the corridor — fifteen cables means losing three or four still leaves capacity on the remainder — worked as designed. Traffic rerouted. Latency increased but services continued. AI training, which typically occurs within a single data center rather than across submarine cables, was not directly disrupted.
The system survived losing four cables. What happens when it loses eight?
The global cable repair fleet consists of approximately eighty vessels. The top five operators — Global Marine Systems, Orange Marine, SubCom, ASN, and Optic Marine Services — control roughly sixty percent of capacity. Sixty-five percent of these vessels will reach the end of their operational lifespan within fifteen years. The industry estimates it needs three billion dollars for replacement ships. At the time of the February 2024 cuts, the Leon Thevenin, docked in Cape Town, was the only repair vessel dedicated to serving the entire African continent.
Eighty ships to maintain every submarine cable on Earth. Sixty-five percent aging out. Three billion dollars needed. This is not a number that suggests the world takes its most vulnerable infrastructure seriously.
The historical parallel is too precise to ignore. In 1902, Britain's telegraph cable network — the "All Red Line" — connected the entire British Empire via cables that landed exclusively on British territory. Britain controlled sixty percent of the world's submarine cables. On August 4, 1914, hours after declaring war on Germany, Britain dispatched the cable ship CS Alert from Dover to cut Germany's five Channel cables. Within hours, four were severed. The goal was to force German communications onto radio, where they could be intercepted. The result was the interception of the Zimmermann Telegram — Germany's secret attempt to ally with Mexico against the United States — which helped draw America into World War One.
In 1911, British strategists estimated it would take forty-nine cable cuts to isolate Great Britain and five to eleven cuts to isolate its major colonies. In February 2024, four cable cuts in the Red Sea disrupted twenty-five percent of Asia-Europe data traffic. More concentrated, more fragile, than the British telegraph network at its imperial height.
The new cables under construction are, in theory, the answer. Meta's 2Africa — the longest submarine cable ever built, at 45,000 kilometers — completed its core system in November 2025, but its critical Red Sea segments remain unfinished, delayed by the Houthi conflict that the cable was partly designed to survive. Google's Blue-Raman system bypasses Egypt entirely for the first time, routing through an overland crossing in southern Israel — an innovation that solves one chokepoint by routing through another contested territory. Its eastern section through the Red Sea is delayed. Arctic cable projects — Far North Fiber from the Nordics to Japan via Greenland and Alaska, Polar Connect from Europe to Asia via the North Pole — promise routes that avoid the Red Sea entirely. They will not be operational before 2027.
Meta has reportedly explored a ten-billion-dollar circumnavigating cable — a single system that would circle the globe and provide true diversified redundancy. The price tag tells you what the market believes: that the problem of geographic concentration in submarine cable routing is worth ten billion dollars to solve.
In the meantime, over ninety percent of Europe-Asia data capacity runs through the Red Sea. The oil tankers and the data cables share the same chokepoint. The Houthis who threaten one threaten the other. Iran, which backs the Houthis and possesses twenty-eight to thirty submarines — including Ghadir-class midget submarines designed for shallow-water sabotage operations and capable of deploying combat divers — has the capability, if not the documented intent, to sever cables deliberately rather than incidentally. The implied threat is itself a weapon. You do not need to cut the cables. You only need the world to know you could.
Seventeen percent of global internet traffic passes through a twenty-six-kilometer gate named for grief. The gate is guarded by no one. The cables inside it are protected by nothing but depth, darkness, and the hope that the next sinking ship drags its anchor somewhere else.