The Anatomy of an Undersea Internet Link

The most important misconception about submarine cables is that the cable itself is the whole system. It is not.

An international fiber route is a chain. Traffic leaves a data center or carrier network.

It moves through terrestrial fiber to a cable landing station. Optical equipment converts traffic onto wavelengths carried by the undersea fibers.

High-voltage direct current feeds electrical power into repeaters installed along the cable. The cable crosses the continental shelf, descends into deep water, may branch toward other landing points, crosses another shelf, and reaches another landing station.

Then traffic enters a terrestrial network again. The undersea portion may be thousands of kilometers long. Yet the critical system includes every point before and after it.

How Submarine Fiber Actually Works

The information is carried as light through optical fibers. Modern systems use multiple fiber pairs, with each pair generally supporting transmission in opposite directions.

Dense wavelength-division multiplexing allows many optical channels to occupy each fiber. The fibers themselves are tiny compared with the cable around them.

Most of the visible structure exists to protect the optical core and deliver electrical power to subsea amplifiers. In deep water, where anchors and fishing gear are less likely to reach, a cable can be relatively slender.

Near shore, where human activity is greatest, cables may be heavily armored and buried beneath the seabed. Repeaters amplify optical signals over long distances.

Branching units allow a trunk cable to split toward multiple landing points. Modern systems increasingly use spatial-division multiplexing: more fiber pairs operating at lower optical power per fiber rather than trying to force ever more power through a small number of pairs.

Meta’s recently completed core 2Africa system uses up to 16 fiber pairs. Its proposed Project Waterworth uses 24 fiber pairs and is designed to span more than 50,000 kilometers across five continents. Those are company-reported specifications, but they illustrate the engineering direction: more fibers; more routes; more flexible optical switching; and enormous aggregate capacity.

Why Cables Fail

The headlines increasingly emphasize sabotage. The statistics still emphasize ordinary maritime activity.

The ITU reported in 2026 that more than 170 cable repairs were recorded worldwide in 2025. Its international advisory work says fishing and anchoring account for more than 80 percent of cable faults, with the remainder typically caused by natural hazards, equipment problems and other causes.

This is important because security policy can become distorted if every cable break is treated as geopolitical attack. An anchor dragged across the seabed can sever a cable.

Fishing gear can snag one.

An underwater landslide can damage multiple systems simultaneously. Earthquakes can shift seabed geometry.

Turbidity currents can travel enormous distances underwater and break multiple cables along a shared route. Equipment can fail.

Shore construction can damage terrestrial segments. Intentional attack is real enough to plan for.

But attribution is difficult.

A suspicious vessel track is not automatically proof of sabotage. A dragged anchor can be negligent, reckless or deliberate.

Investigators may need ship records, seabed evidence, cable damage patterns, crew testimony and intelligence before assigning intent. That distinction should remain explicit.

Why the Internet Usually Survives a Cut

The global Internet was built around routing. Traffic does not have to follow one permanent physical path.

If one cable loses service, carriers and cloud networks can redirect traffic through another system. That is why a dramatic cable fault may produce only elevated latency rather than an obvious outage for ordinary users.

When multiple Red Sea cables were disrupted in September 2025, Microsoft said Azure traffic could be rerouted through alternate paths. Users could experience higher latency, but the affected route did not simply erase Azure connectivity.

This is the strength of a mesh of networks. But rerouting has limits.

Alternate routes need spare capacity. They may add geographic distance.

That increases latency.

Several alternate routes may pass through the same secondary chokepoint. If multiple cables fail simultaneously, remaining paths can become congested.

And some places have very little redundancy to begin with. The ITU has repeatedly emphasized that small island developing states and underserved regions can be highly dependent on one or a few systems. This produces a critical distinction: global Internet resilience does not guarantee national or regional Internet resilience.

The Chokepoint Problem

Ocean-scale networks are constrained by geography. Europe and Asia are separated by enormous distances.

Island chains create natural landing corridors. Shallow seas concentrate maritime traffic.

Political borders and permitting shape where cables can land. Cloud regions and data centers create demand around particular cities.

The result is a network with hubs.

Hubs are efficient.

They are also concentration points. A strategic chokepoint can exist at several levels:

a narrow sea;

a coastal landing cluster;

a terrestrial crossing;

a cable station;

a limited set of repair ports;

or a regulatory jurisdiction through which multiple systems must pass. This is why a map showing dozens of lines can create false comfort. Lines that appear separate in the middle of an ocean may converge at exactly the place most exposed to anchors, construction or politics.

The Red Sea and Egypt Corridor

No major cable route better demonstrates the difference between global redundancy and geographic concentration than the Europe-Asia corridor through Egypt and the Red Sea. CSIS estimated in a 2025 case study that roughly 17 percent of global Internet traffic and more than 90 percent of Europe-Asia communications traverse Egypt-linked subsea systems.

The exact share changes as new systems enter service and routing shifts, so those figures should be treated as an estimate rather than a permanent constant. The structural point is durable.

Egypt occupies the shortest practical bridge between the Mediterranean and Red Sea. Cables land on one coast, traverse Egypt terrestrially or through managed infrastructure, then re-enter subsea systems on the other side.

At the southern end of the Red Sea, traffic passes near the Bab al-Mandab corridor. That creates layered exposure:

dense shipping;

shallow-water cable routes;

regional conflict;

insurance risk;

repair-vessel access constraints;

and concentration of multiple systems in the same geographic space. The September 2025 Red Sea cuts demonstrated the practical effect.

Internet users across parts of the Middle East and South Asia experienced disruption or degraded performance. Microsoft rerouted Azure traffic but warned of higher latency.

The cause of those specific cuts was not immediately established. That uncertainty matters. The resilience problem exists regardless of whether a particular failure is accidental or hostile.

The Baltic Security Laboratory

The Baltic Sea has become the world’s most visible laboratory for protecting undersea infrastructure in a gray-zone security environment. Multiple telecommunications and energy cables were damaged across 2023, 2024 and 2025, producing investigations involving commercial ships, anchors and possible intentional acts.

NATO launched Baltic Sentry in January 2025 following damage to critical infrastructure between Finland and Estonia and other regional incidents. The operation added frigates, maritime patrol aircraft and naval drones to improve awareness and deterrence.

By February 2026, NATO allies moved further. Task Force X-Baltic transitioned from experimentation toward nationally owned capabilities that NATO could integrate for more persistent maritime surveillance. NATO says its 2025 demonstrations involved 70 air and maritime drones, with AI-enabled systems contributing to situational awareness around critical undersea infrastructure.

The European Union is building a parallel civilian and regulatory response. In June 2026, the European Commission funded its first regional cable hubs in the Baltic and Mediterranean and launched a €40 million call aimed at expanding repair capability.

The strategy is notable because it does not assume that every cable can be physically guarded. The Baltic contains too much maritime traffic and too much infrastructure for continuous close protection of every kilometer.

Instead, the emerging architecture combines: maritime domain awareness; AIS and vessel intelligence; autonomous sensing; military presence; industry coordination; repair readiness; route redundancy; and legal enforcement. That is a more realistic model of infrastructure security than imagining a warship stationed over every cable.

Landing Stations and Shallow-Water Vulnerability

The deepest part of the ocean may be the safest part of a cable. Fishing equipment rarely reaches abyssal depths.

Anchors do not drag across most deep-ocean cable routes. The cable becomes physically more exposed as it approaches land.

That is why modern systems use heavier armor and deeper burial in high-risk shallow-water zones. Meta says Project Waterworth will maximize deep-water routing and use enhanced burial techniques in areas more exposed to anchors and other hazards.

Its 2Africa project likewise used increased burial depth and route engineering around known seabed risks. But landing-zone resilience is not only about the cable in the surf.

A landing station contains power feeds, optical terminal equipment, monitoring systems and connections to terrestrial networks. Several cables may enter the same facility or nearby facilities.

From there, terrestrial backhaul may follow common ducts, railroad corridors, roads or metro fiber routes. A country can therefore possess multiple international cables and still have a terrestrial single point of failure. True cable diversity requires: different shore approaches; different landing stations; different terrestrial routes; different power dependencies; and ideally different regional paths.

The Repair-Ship Bottleneck

One of the least glamorous components of the Internet may be one of the most strategically important: the cable repair ship. When a cable fails, engineers first localize the fault electrically and optically.

A repair vessel must then travel to the area. The crew grapples for or retrieves the damaged cable.

The damaged section is brought aboard. Technicians splice new fiber and conductor sections.

The cable is tested.

Then it is returned to the seabed and, where appropriate, buried again. That operation can be routine in calm conditions and extraordinarily difficult in conflict zones, storms, deep water or constrained jurisdictions.

Repair time is affected by ship availability, distance to the fault, weather, permits, customs, security, insurance, spare cable and repeater availability, and whether several faults are competing for the same vessels. The ITU’s 2026 international resilience work specifically identified limited repair-vessel availability in some regions and increasing repair delays as a global weakness.

This is why the European Commission’s new repair-capacity funding matters. The Internet cannot be considered resilient merely because redundant cables exist. If several fail and the restoration queue stretches for months, resilience steadily decays.

Google, Meta and the Hyperscaler Cable Era

Historically, submarine cable systems were dominated by telecommunications carriers and multinational consortia. Cloud computing changed the economics.

Companies such as Google and Meta now move extraordinary quantities of traffic between their own data centers, cloud regions, content caches and users. AI is increasing that demand again.

Google said in 2026 that its infrastructure includes roughly 10 million kilometers of terrestrial and subsea fiber connecting more than 30 data centers and more than 40 cloud regions. Meta’s Project Waterworth is a multi-billion-dollar proposal spanning more than 50,000 kilometers and five continents.

Meta’s 2Africa system now links East and West Africa with Europe, the Middle East and South Asia, with the core infrastructure completed in 2025 and extensions continuing. Google has been building new Pacific and Indian Ocean routes deliberately designed to avoid existing concentration points.

Its TalayLink project, announced in late 2025, is planned to connect Australia and Thailand over a route west of the Sunda Strait, creating a more diverse path than many existing systems. This shift is strategically important.

The companies that build AI models and cloud platforms increasingly own part of the physical infrastructure connecting the regions where those models run. The AI stack therefore extends farther downward than a data center.

It extends through terrestrial fiber, landing stations, ocean routes, repair agreements and geopolitical permissions. That is another example of SURVXCOM’s central technology thesis: software power rests on physical systems.

Supply Chain, Ownership and National Security

A submarine cable can be secure physically and still raise national-security questions. Who built the repeaters?

Who owns the landing station?

Who has maintenance access?

Which repair company handles a fault? Who controls network-management systems?

Which jurisdictions can compel access? Which entities lease capacity?

The United States has tightened its regulatory approach around those questions. In 2025, the FCC adopted changes to submarine-cable licensing intended to address foreign-adversary ownership and security risks while modernizing a process that had not undergone a comparable comprehensive revision for decades.

The FCC’s action created new reporting and foreign-adversary restrictions and pursued additional measures around trusted equipment, physical security and cybersecurity. Members of Congress separately pressed major U.S. technology companies for information about foreign repair firms, equipment and suspected irregularities.

These concerns should be handled carefully. A foreign-owned repair vessel is not proof of espionage.

A Chinese-manufactured component is not proof that data has been compromised. But submarine cables are infrastructure through which strategically important data flows. Governments therefore increasingly treat ownership, supply chain and maintenance access as part of the attack surface.

Can Satellites Replace the Cables?

No—not at comparable global capacity. But they can change the resilience equation.

LEO satellite networks can provide alternate connectivity to islands, ships, remote regions and disaster zones. They can preserve communications when a terrestrial landing point fails.

They can support emergency backhaul. They can reduce the consequences of losing a single cable.

But Article 008’s central distinction applies again: coverage is not capacity. A dense bundle of fiber pairs can transport enormous volumes of traffic with low marginal energy per bit.

Satellite capacity must be shared across radio spectrum, orbital beams and gateways. Even space networks frequently depend on terrestrial fiber once traffic reaches a gateway.

The future is therefore hybrid:

When the Cable Becomes a Sensor

The fiber itself may become part of its own protection system. Distributed acoustic sensing uses changes in backscattered light inside optical fiber to detect vibration along the cable.

Research published in 2026 demonstrated a labeled dataset using distributed acoustic sensing along a North Sea cable segment to detect and estimate the distance of vessels relative to the cable. Other recent experiments have used operational telecom fibers to detect earthquakes, ocean waves and even tsunami signals.

This is an emerging field rather than a universal deployed security system. But it suggests an intriguing evolution:

the cable may no longer be only a communications pipe. It may also become a distributed environmental and security sensor stretching for kilometers across the seabed.

The challenge will be separating useful signal from noise. A cable can detect vibration.

That does not automatically establish malicious intent. Sensor data must be combined with vessel tracking, maritime intelligence, seabed mapping and human investigation.

The SURVXCOM Subsea Resilience Test

Counting cables is not enough. A resilient international network should answer twelve harder questions.

What to Watch Next

The Internet Beneath the Water

The modern Internet is often described as virtual. It is not.

It is glass.

Copper.

Steel.

Electricity.

Ships.

Ports.

Landing stations.

Permits.

Seabed routes.

Repair depots.

Cloud networks.

And agreements between countries.

The magic of the Internet is that users rarely have to think about any of it. A video call from Charlotte to London feels instantaneous.

An AI model in one continent can serve a user in another. A financial institution can move data across oceans in milliseconds.

That experience is possible because enormous physical systems work quietly beneath it. Submarine cables are one of the clearest examples of a broader principle running through the entire Critical Technology Stack:

digital power is physical power disguised by a good interface. The world is adding more cables.

More fiber pairs.

More routes.

More landing sites.

More satellite alternatives.

That should improve resilience.

But only if the new infrastructure reduces shared failure domains rather than merely adding capacity along the same corridors. The important question is not:

How many submarine cables exist?

It is:

How many truly independent ways does the data have to get there? That is the architecture beneath the global Internet.