Undersea Cables: The Achilles' Heel of AI

The Precedent: 1914, When Cables Decided a War

In 1914, hours after the declaration of war, the British ship Telconia sailed into the North Sea and cut the five telegraph cables connecting Germany to the world. The mechanism was simple: control the flow. Berlin lost direct access to its embassies and markets, while London read the remaining traffic. Whoever controlled transmission also controlled interpretation. Germany discovered too late how costly that cut would prove: every communication had to pass through slower, more exposed channels, often under British surveillance.

A century later, the same mechanism remains active. The context is different; the structure is identical.

Whoever controls the cables controls information. And today, information encompasses the entire AI economy, from trained models to inferences distributed across continents. The difference from 1914 is the stakes: then it was diplomatic dispatches; today it is flows that power entire productive sectors.

The Current Pattern: AI Rests on Underwater Pipes

Nearly 100 percent of intercontinental internet traffic passes through cables just a few centimeters wide. This is documented in an analysis published by Lawfare^[1], which also quantifies speed: information travels five times faster via cable than via satellite. Satellite is not an equivalent alternative. It is a slower fallback, with lower capacity and higher latency.

Around 600 cables support the planet's digital backbone. They are tubes. Fragile, exposed, contested.

AI amplifies every dependency. Every transnational data center rests on these underwater wires, and the demand trajectory is accelerating faster than laying capacity. A model trained on one continent and served on another crosses these cables with every query. The U.S. plan for AI dominance lists 90 recommendations and overlooks the roughly 600 backbone cables that make everything else possible. It is a structural oversight: computing power is planned while the channel connecting it to the rest of the world is taken for granted.

The Bottleneck: Licenses, Timelines, National Capacity

The real brake is administrative. The Federal Communications Commission has proposed a rule to accelerate cable license reviews. A step in the right direction.

The current system clogs projects. A review designed to take 120 days often stretches to six or eight months, according to a participant cited in the same analysis. Private actors demand predictable timelines before committing capital to costly infrastructure. A multi-month delay is not merely a bureaucratic inconvenience. It shifts the return calculation of a multi-year project, and sometimes kills it before it begins.

Demand far exceeds capacity. Around 100,000 miles of new cables are needed by 2040 to handle anticipated traffic, and the proposed rule falls well short of that leap. Accelerating licenses does not resolve the bottleneck if ships, shipyards, and skilled labor are lacking. Improving cable quality and monitoring the seabed remain chapters still open.

The Mechanism: How Fragility Becomes Leverage

Vulnerability transforms into an instrument of power. Three forces erode the system's resilience in parallel.

• Currents, marine creatures, and ordinary wear degrade cables every year.
• Hostile actors add intent to natural degradation.
• The ocean floor receives very little continuous surveillance.

The combination matters more than the individual threats. A poorly monitored seabed makes a deliberate act indistinguishable from an accidental failure. Whoever cuts a cable can deny it, and the denial holds as long as surveillance is absent.

The International Cable Protection Committee collects failure data and coordinates protection among operators, as described on its portal (iscpc.org^[2]). Its records show hundreds of repairs per year, most linked to fishing and anchoring. This figure cuts both ways: the majority of failures are ordinary, but precisely that background noise covers intentional action.

The Trump administration has lowered barriers to deep-sea mining. More activity on the seabed means greater exposure for cables. The geography of digital power is shifting toward those who patrol the seas, and this rewrites the map of infrastructure risk.

My Position

Here is the thesis in one sentence: national capacity in undersea cables is a first-class strategic asset, as undervalued as energy reserves were in the 1970s.

Markets treat cables as boring utilities. They are, in fact, the single point of failure of the digital economy. Their fragility remains the worst-priced infrastructure risk of the decade.

What would change this position? A coordinated program among the United States, the European Union, and Japan to double annual cable laying and install permanent seabed surveillance. Concrete evidence of that coordination would reduce my conviction. At the moment, the trajectory points in the opposite direction.

Three Implications for Capital

First, 36-month horizon. Family offices and sovereign wealth funds should study the owners of cable-laying ships and repair companies. The global fleet is aging while demand explodes, and the bottleneck generates lasting pricing power.

Second, 24-month horizon. A CEO with transnational data centers carries geopolitical risk still absent from most board strategic plans. Redundancy in digital routing deserves a dedicated budget line, separate from the cloud chapter.

Third, 18-month horizon. A Chief Risk Officer should insert the coordinated cable-cutting scenario into VAR models. Today that scenario remains outside most simulations, and its absence distorts every measure of operational continuity.

The Forecast

The market still misprices cable risk. Here is the explicit forecast, with horizon and verification indicator.

By December 2026, at least one new incident of deliberate damage to undersea cables will be documented in NATO waters. Confidence: 78 percent. Horizon: December 31, 2026. Verification: an official report from an allied government or the ICPC attributing the damage to intentional action.

The falsification signal remains clear: zero documented deliberate incidents by that date.

What to Watch

Three precedents are enough to call it a pattern. The Baltic has already provided them over the past few winters.

• The final text of the FCC rule and actual average authorization timelines.
• New cable-laying ship orders and the shipyards receiving contracts.
• ICPC and allied government reports on failures attributed to deliberate action.

This is a regime change in infrastructure security, a structural shift, distinct from a passing cycle. Those who price it now will be buying scarcity at a discount.