Hook: Today’s Abyss Digest slipped in a line a normal person would scroll past: "The Houthis are bringing war back to Yemen… a series of strikes on the government port of al-Makha… the Saudi Aramco refinery in Jizan… a commercial vessel in the Bab-el-Mandeb Strait." And further down: "Iran’s ‘Axis of Resistance’ received orders to prepare for the closure of Bab-el-Mandeb." That "further down" is the most interesting part. Because Bab-el-Mandeb isn’t just a narrow strait between Yemen and Djibouti, through which 12% of global maritime trade passes. It’s also where 17 underwater fiber-optic cables carry 17% of global internet traffic and over 90% of Europe-Asia data flows, all squeezed through a 26-kilometer bottleneck. And when, in February 2024, the Houthis hit the bulk carrier Rubymar (171 meters long, 21,000 tons of fertilizer on board) with a ballistic missile, and the crippled ship drifted for two weeks with its anchor dragging—it apparently snagged three of those cables. By accident. Not intentionally. Just because a 30-mile drift of a burning ship loaded with fertilizer, in a zone where cables lie 1.5 kilometers apart, statistically ends exactly how it should. And here’s what hit me when I dug deeper: in this story, there isn’t a single villain. There’s a poorly aimed $50,000 missile, there’s the current in Bab-el-Mandeb, there’s a 26-kilometer geographic funnel, there are three cables lying on the seabed 700–1,000 meters apart, and there’s a BlackSky satellite image showing the anchor of a dead ship slowly plowing the bottom of the Gulf of Aden. And that anchor—a 30-meter chain with a 10-ton weight—turns out to be the only physical object that simultaneously knocked out 25% of Europe-Asia internet traffic. Not AI, not a state hacker, not a nuclear strike. One steel hook on a chain, dangling from a drifting ghost.
And while the world debated the Houthi missile strike and insurance for tankers, something happened on the seabed that Wired, AP, Bloomberg, and Kentik independently confirmed in April 2024: the timestamps of the three cable outages—Seacom/Tata (9:46 AM on February 24), AAE-1 (9:51 AM, +5 minutes), EIG (already partially faulty)—matched the predicted drift trajectory of the Rubymar. Doug Madory from Kentik, who analyzed BGP traffic in real time, saw how East African providers—Tanzania, Kenya, Uganda, Mozambique—lost connection in 5 minutes. And then the disruption rippled to Vietnam, Thailand, and Singapore. Three underwater cables serving millions of people lie on the seabed because one crewless ship slowly drifted with the current. This isn’t sabotage. It isn’t a covert operation. It’s entropy in its purest form.
Bab-el-Mandeb (Arabic: باب المندب, "Gate of Tears" or "Gate of Sorrow"—yes, that’s really what it’s called; the name comes from the danger of navigation) is a 20-mile (26-km) strait between Yemen to the north and Djibouti to the south, connecting the Red Sea to the Gulf of Aden and onward to the Indian Ocean. It’s one of two points through which communication between the Mediterranean Sea and the Indian Ocean passes (the other is the Suez Canal, 193 km to the north). If you want to send a bit from London to Mumbai via the shortest route, it goes right here.
And here’s the key: through this funnel, 17 active underwater cables physically pass (other estimates say 12–16, depending on whether you count cables running through the Red Sea but not the narrowest part; see Subsea Cables, June 2026). Each is a bundle of glass fibers in a sealed polyethylene and copper sheath, about the thickness of a garden hose, laid at depths of 150–1,500 meters on the seabed. They all pass through the same 26-kilometer bottleneck because there’s no other way. Going around Africa means +15,000 km and +60 ms of latency. Going through Central Asia means terrestrial hops with no equivalent capacity. Going through the Arctic isn’t feasible yet. So 17 cables packed into a 26-kilometer geographic channel isn’t a design flaw—it’s the geometry of the Eurasian-African continent.
Wired, in April 2024, citing TeleGeography, emphasized: "Twelve of the cables run through the Bab-el-Mandeb Strait… These cables vary massively in their age, also in their capacities" (Alan Mauldin, TeleGeography). Together, these cables carry 17% of global internet traffic and over 90% of Europe-Asia data flows (WANA, July 25, 2026; IAI, December 2025). To put the scale in perspective: in this 26-kilometer strip of water, there’s more internet capacity on the seabed than in the world’s 50 largest data centers combined. And it’s all in a zone where, on the surface, a real war is happening with ballistic missiles, drones, and dead ships.
TeleGeography, in September 2025, published a figure: 1,835 Tbit/s—that’s the total capacity of all international internet connections on the planet (TeleGeography, September 23, 2025). Seventeen percent of that is about 310 Tbit/s, theoretically passing through Bab-el-Mandeb. The actual working cables—about 12—together carry, by various estimates, 80–110 Tbit/s of active traffic. That’s equivalent to ~80 million simultaneous HD video calls. And when the Rubymar drifted, at some point—literally—a quarter of that flow dropped.
To understand how this was not a targeted attack but entropy in real time, you need to walk through this timeline step by step. Wired, in April 2024 (Matt Burgess, "A Ghost Ship's Doomed Journey Through the Gate of Tears"), reconstructed it using satellite imagery, AIS data, and interviews with maritime analysts. I’ll retell it with engineering precision.
February 18, 2024, ~8:00 PM local time. Two opposing missile strikes (per CENTCOM—anti-ship ballistic missiles) hit the Rubymar in Bab-el-Mandeb. The crew of 24 (11 Syrians, 6 Egyptians, 3 Indians, 4 Filipinos) sends a distress signal. At 9:30 PM, the Rubymar transmits its last AIS signal. By 1:57 AM on February 19, the crew is rescued by the container ship Lobivia and coalition vessels led by the U.S. After the rescue, the crew dropped anchor—standard procedure for a sinking ship. Then the vessel was abandoned.
February 19–22. The Rubymar, with its anchor down, drifts north with the current and wind. BlackSky satellites track its position: 30 nautical miles (~55 km) north of the missile impact point. An oil slick trails behind it. The depth in this area is about 150 meters (Wired, citing Kentik). That’s an important number: 150 meters is a zone where the anchor could snag a cable (typical depth for trunk cables on the continental shelf).
February 22, ~2:00 PM. BlackSky satellite images show a mysterious circular wave pattern around the Rubymar. A former naval analyst (name withheld for security reasons) tells Wired: "this could be a sign the anchor may have come loose." In one image, he says, there’s an "unidentified object, which could be a small boat, nearby"—meaning a possible vessel that might have cut the anchor chain. But there’s no proof—just interpretation.
February 23–24. Wind and current intensify. The Rubymar starts drifting faster. Robert Parkington from Geollect (a geospatial analytics firm) says: "As wind increases, as current increases, that chance for movement gets so much higher… Even a small breeze can have an impact on where the vessel's moving." February 24, 9:46 AM UTC—Seacom/Tata cable (15,000 km, East Africa ↔ India) loses connection. Five minutes later, at 9:51 AM, AAE-1 (25,000 km, Europe ↔ East Asia) goes down too. The third cable, EIG (15,000 km, UK ↔ India), was already partially faulty (Doug Madory, Kentik).
March 2. The Rubymar finally sinks—for good this time. On board: 21,000 tons of fertilizer, now slowly leaching into the water. An environmental disaster, separate from our story.
Total: 12 days from missile impact to ship sinking. During those 12 days, a dead ship with a steel anchor dangling from a 30-meter chain drifted across a seabed pre-laid with cables. And when they met—the cables broke. Not because the Houthis were aiming at the cables. But because the seabed of Bab-el-Mandeb is as much a war zone as its surface, and the seabed has no Minsk agreements of its own.
Here’s where the most underrated part begins: when the cables went down, the internet didn’t crash. Microsoft Azure confirmed disruptions starting February 26—but these were increased latencies, not outages. PTCL (Pakistan) reported reduced capacity. Riyadh recorded packet loss. In East Africa, providers started switching to backup routes. And all this happened in 5 minutes.
How does this work? BGP routing automatically reroutes traffic to surviving cables. SEA-ME-WE 5 (24 Tbit/s)—the "only major surviving cable in that corridor"—took on the main load. AAE-1 (40+ Tbit/s, though partially damaged) also kicked in. Some traffic went through Central Asia via terrestrial networks (through Russia, Kazakhstan, Uzbekistan), some around Africa via the Cape of Good Hope, adding ~60 ms of latency to the Singapore–Marseille route. Latency in Asia and Europe increased by 20–40 ms for many routes.
This is the architectural miracle of redundancy that engineers have built over 30 years. 400+ underwater cables worldwide (JRC, NOAA, June 2024). The BGP protocol, developed in 1989, instantly reroutes paths. Every IP packet is essentially a miniature GPS query: "Where’s the nearest exit?" And the answer is dynamic, with no central control.
GeoCables, a monitoring company, analyzed real traceroute data from Minsk, Almaty, and Tbilisi. They found that even after cable repairs, BGP routes didn’t return to their previous paths: "Traffic that previously entered the Middle East network at Jeddah or Aden began appearing more frequently at alternative entry points—reflecting carriers' decisions to diversify away from Yemen-proximate landing stations" (GeoCables, January 2026). Providers have "remembered" the hit for years to come and are keeping backups. This is infrastructure with memory—like an immune system that learns from every infection.
But the cost of this architectural miracle is 60 ms of latency, 20–40 ms of quality loss, thousands of extra kilometers on routes, plus financial losses: the affected cables carried about 25% of Europe-Asia traffic, and rerouting required purchasing capacity on alternative cables. Alan Mauldin from TeleGeography comments in November 2025: "They are not only unable to monetize their investments by sending data over these cables, but they are forced to purchase capacity on alternative cables to meet their near-term requirements" (Insurance Journal/Bloomberg, November 17, 2025). So not only is traffic lost—money invested in these cables is lost too. And owners are starting to sell them at a discount: Aqua Comms (Ireland), owner of EMIC-1 and part of 2Africa, sold the company at a discount in January 2025, directly citing in documents: "indefinite delay to EMIC-1… due to ongoing conflicts in the Red Sea".
Here’s where the real drama begins—the reason I dug into this topic. Restoring cables in an active conflict zone is not a technical problem but a political one, and in this game, everyone loses.
AAE-1 (Asia Africa Europe-1) was repaired only in July 2024—five months after the break. And in an IAI brief, Giacomo Leccese (December 2025) explains why it took so long: "Authorisation to repair the AAE-1 cable was granted only in July, following months of negotiations between operators and the two Yemeni authorities" (IAI). Yemen has two governments—the internationally recognized one in Aden and the Houthi one in Sana’a. Both demand permits. And both procedures are toxic for the operator.
The Aden government denied permits because the AAE-1 consortium includes TeleYemen—a telecom operator linked to the Sana’a administration, i.e., the Houthis. Any repair permit would mean de facto recognition of TeleYemen’s role in international infrastructure. And in November 2025, insurers note that "the US designates the militia as a terrorist organization," meaning any transaction with the Houthis could violate American sanctions (Curtis Client Alerts, March 25, 2024).
The Houthis, in turn, declared: "all submarine cable ships must first obtain a permit from the Maritime Affairs in Sana'a before entering Yemeni territorial waters" (Mosfer Alnomeir, post on X, March 4, 2024). That is, they demand that any vessel in Yemeni waters must first get permission from the Houthis. And this is a legal dead end: no state in the world recognizes the Houthis as Yemen’s legitimate authority, and the U.S. officially considers them terrorists. Making a deal with the Houthis violates sanctions. Not making a deal means no access to cables lying in their territorial waters.
And here’s the kicker: the Houthis deny involvement in damaging the cables. "Houthis Deny Targeting Underwater Cables Amid Marine Disaster Warning" (The Guardian, February 28, 2024). They say it was the Rubymar’s drift, not their targeted attack. And this position is internally consistent: if they admitted to hitting the cables, they’d de facto justify their own strikes on ships (protecting critical infrastructure). As it stands—no casus belli for the international community. This is legal-narrative acrobatics of the highest order.
Cable operators broke the deadlock after six months of negotiations, and insurance for cable ships near Yemen jumped to $150,000 per day (Wall Street Journal, March 3, 2024). That’s five times higher than standard cable ship insurance. And even after repairs, the insurance didn’t drop—because the political instability in the region didn’t go away.
In November 2025, Bloomberg dropped a reporting bomb: "Google, Meta Delay Red Sea Cables as Security Risks Rattle Plans." Here’s what they found. Five major new cables meant to pass through the Red Sea are frozen indefinitely:
That is, the very corporations that own the world’s internet infrastructure can’t expand it in the zone where it’s needed most. And they’ve started looking for other routes—and these routes are geopolitically charged.
Option 1. Through Saudi Arabia and Bahrain. Previously considered "too expensive and indirect." Now—"viable." Terrestrial fiber through the desert, across borders, connecting to an underwater cable in the Persian Gulf, completely bypassing Yemen. That’s +500 km of overland route, but politically—much more stable.
Option 2. Through Iraq. "Even a route through Iraq, historically seen as geopolitically risky, has become a possible alternative" (Asoz Rashid, IQ Group). E&, Ooredoo, Gulf Bridge International have already bought capacity on the Silk Route—a cable system through Iraq, Iran, Turkmenistan, Kazakhstan. The Iranian corridor, which used to be a sanctions minefield, is now a potential lifeline for the global internet. The irony is worthy of a fable.
Option 3. Direct negotiations with the Houthis. Bloomberg notes that some operators are considering requesting an exemption from the U.S. Treasury Department—a direct carve-out from sanctions allowing talks with the Houthis. And NATO is on this list—operators are considering calling on the alliance for help. That is, NATO, created to protect the transatlantic cable in the 1950s, may be drawn into protecting the cable between Europe and Asia in 2026. Not as a military alliance, but as an infrastructure alliance.
And here, finally, is the legal layer that struck me as an engineer more than anything else in this story. Underwater cables have been protected by international law since 1884—the Convention for the Protection of Submarine Telegraph Cables (Paris, March 14, 1884, revised in 1958 and in 1982 as UNCLOS Part VII). Tara Davenport, an assistant professor at the National University of Singapore, analyzed the applicability of existing norms to modern attacks in October 2023 for Lawfare (“Intentional Damage to Submarine Cable Systems by States”).
And here’s what she found: the existing legal framework was written for the telegraph era, and it hasn’t kept up with the fiber-optic age. Specifically:
The 1884 Convention criminalizes "willful or negligent" damage to cables, but only in areas beyond territorial waters (i.e., outside the 12-nautical-mile zone—inside that, it’s national jurisdiction). And in 1884, the "territorial waters" problem was simple—3 miles from shore. Now it’s 12 miles, and all Bab-el-Mandeb cables pass within or near the 12-mile zones of Yemen and Djibouti. So formally, the 1884 Convention’s protection doesn’t apply to them.
UNCLOS (United Nations Convention on the Law of the Sea) 1982 recognizes the freedom to lay underwater cables (Article 79), but doesn’t establish criminal liability for damaging them. That remains a matter of national legislation.
"Use of force" in international law (jus ad bellum)—after Nord Stream in 2022, there’s active debate over whether sabotage of underwater infrastructure constitutes an "armed attack" giving the right to self-defense under Article 51 of the UN Charter. EJIL Talk! published an analysis (“Are sabotage of submarine pipelines an ‘armed attack’”), and there’s no consensus. If an underwater cable is critical infrastructure, does destroying it = "armed attack"? Or is it a "gray zone"—below the threshold of armed attack but above ordinary crime?
And here’s the key: the nature of the damage in Bab-el-Mandeb isn’t a "gray zone." It’s an accident. If the Houthis had intentionally cut the cables, that would be a "gray zone." But here—a dead ship drifts with the current and accidentally snags three cables with its anchor. And there’s no legal framework for this case. No article, no precedent, no state to hold accountable because there’s no aggressor state.
And here I want to take a breath and say the main thing that struck me in this story. Bab-el-Mandeb isn’t a military crisis dressed up as a technical one. It’s a technical crisis that the military made inevitable.
What I mean is this. Underwater cables were designed in an era when the seabed was a zone of peace. When the transatlantic telegraph cable was laid in 1858/1866, no one thought it would be cut by ballistic missiles. No one thought a state would refuse to issue repair permits because it has two governments, both illegitimate in the eyes of the international community. Cables were designed with the assumption that "the seabed is neutral." That’s the same assumption as ARINC 429 in the Boeing 737 (see my recent curiosity about the ESP32)—"physical access is impossible." And just like in Boeing, this assumption turned out to be false in the long run.
And here’s what hit me in this story. In 1884, at a conference in Paris, 28 states gathered to agree on protecting underwater cables. It was the first international treaty on critical infrastructure in history. And it worked for 140 years—for the telegraph, then coaxial cables, then fiber optics. 140 years on the same legal framework.
And in 2024, one drifting ship with a broken anchor in 12 days showed that this framework doesn’t cover cases that are physically possible in 2026. Not because the framework is bad—but because the world it was designed for is over.
The most striking thing about this story is how much we now know about the geography of the digital world, and how little we know about its vulnerability.
What we know:
What we don’t know:
You know what struck me most in this story? Not the Houthis, not the missiles, not the Rubymar’s drift, not BGP redundancy, not the legal labyrinth of 1884. It was one line from Wired that I couldn’t forget afterward: "There are also several types of anchor… If the soil type is not right, and the cable has quite shallow burial or it is on the seabed, you are going to catch it if your anchor starts to drag" (Michael Brown, University of Dundee, 2024).
That’s one sentence—and it describes the entire architecture of modern civilization’s fragility. Cables aren’t buried deep. An anchor dangles in the water. The seabed is the wrong type. And that’s enough.
When I started digging, I thought this would be a story about the Houthis—geopolitics, the Middle East, military conflict. Instead, it turned into a story about physics: one steel hook on a 10-ton chain, dangling from a drifting ship in a zone where 17 cables lie 700–1,000 meters apart. Statistics work without malice. The anchor doesn’t know it’s cutting the internet. The cable doesn’t know it’s being cut. The BlackSky satellite captures it happening, and no one can stop it.
And here’s what seems most important to me. This story isn’t over. In 2024, the Rubymar drifted for 12 days and accidentally snagged three cables. In 2025, five new cables through the Red Sea are frozen. In 2026, insurance for cable ships off Yemen costs $150,000 per day. Repairing AAE-1 took 5 months, and none of the major publishers have publicly guaranteed it’ll be faster next time. And the Iranian corridor—through a sanctioned country that Western telecom operators haven’t dealt with for decades—is already being discussed as a realistic detour. That is, the world hasn’t solved the Bab-el-Mandeb problem. The world is bypassing it. And that’s different.
And in this, I think, lies the deepest engineering lesson of 2024, which we somehow missed while obsessing over missile strikes and insurance for tankers. When critical infrastructure runs into a 26-kilometer geographic funnel in an active conflict zone, there’s no engineering solution. Only a political one. And if there’s no political solution—all that’s left is a detour. And the detour goes through Iran.
Which, you’ll agree, isn’t the plot twist we planned in 1884 in Paris, when 28 states signed the Convention for the Protection of Submarine Telegraph Cables.
📚 Sources: