The 99-percent chokepoint.

Submarine cables, the shadow-fleet doctrine, and what redundancy by diversity actually requires. A study on the most consequential infrastructure most users have never thought about.

Prologue. Where the internet actually lives.

Approximately 99 percent of intercontinental data traffic moves over submarine cables. Satellite links, including the rapidly maturing low-earth-orbit constellations, account for a small share of long-haul international bandwidth and a smaller share of latency-sensitive transit. The cloud, in the only sense the word has operationally, is a thin glass strand on the seafloor. There are roughly 574 active cable systems globally as of April 2025, totalling on the order of 1.4 million kilometres of glass. They are owned by a small set of consortium investors and, increasingly, by hyperscaler cloud operators. They are protected, in international law, by a treaty of 1884.¹

The 1884 Convention for the Protection of Submarine Telegraph Cables remains the operative instrument. The substance is reasonable for a 19th-century commerce-protection instrument and inadequate for a 21st-century strategic-asset-protection instrument. The cables themselves have evolved several orders of magnitude in capacity. The legal architecture around them has not.

The thesis.

A 99-percent dependency is by definition a chokepoint. The architectural countermeasure to a chokepoint is redundancy along a different axis from the dependency. For submarine cables that means more cables, more diverse landing points, more diverse owners and more diverse protective regimes. None of this is technically controversial. All of it is institutionally hard, and the institutional hardness is the entire reason the chokepoint persists.

The shadow-fleet incident series of 2023 to 2025 has made the dependency operationally legible. It has not, so far, produced an architectural response at the scale the dependency warrants. This study sets out the dependency in numbers, the incident series in chronology, and the architectural response that the libertarian read of resilience would prescribe.

The numbers.

Cable systems are typically owned by consortia of telecommunications carriers, increasingly with hyperscaler cloud operators (Google, Meta, Microsoft and Amazon Web Services) anchoring or directly owning new systems. New-build capacity through 2026 is overwhelmingly hyperscaler-anchored on the trans-Atlantic and trans-Pacific arteries. The trans-Atlantic Marea cable, jointly owned by Microsoft and Meta, has been a milestone for that ownership shift; the trend has accelerated since.¹

The shift from carrier-consortium to hyperscaler ownership is the most consequential structural change in the cable industry of the past decade. It has implications the policy literature has not yet caught up with. A trans-oceanic cable owned by a US-headquartered hyperscaler is, in legal effect, US infrastructure. The German, French or Spanish state has limited strategic standing over the operation of a cable that lands in its territory but is owned by a Mountain View parent company. The standing exists in landing-station regulations and emergency-cable powers; it does not exist in the cable's commercial governance. Whether this is good or bad depends on which alignment one is measuring; that it is structurally different from the prior regime is not in dispute.

The Baltic theatre.

The Baltic Sea has been the most-reported submarine-cable theatre of the past 30 months. The basin is shallow, narrow and densely instrumented, which means that incidents occur with relative frequency, and the detection-to-public-attribution latency is short by maritime standards. The series of suspicious incidents listed below is composite and does not exhaust the public record; it is the cluster on which Western strategic discourse has converged.⁴

The doctrinal point is unaffected by the legal classification of any individual incident. Whether each event is, in court, an accidental anchor drag or a deliberate sabotage operation is materially less important than the fact that the cumulative pattern functions as deterrent signalling regardless of the individual classification. The shadow fleet does not need a single confirmed attribution. It needs a credible pattern.³

The repair-time problem.

Submarine cable repair is performed by a small global fleet of specialised cable ships, on the order of 60 vessels worldwide. The fleet is disproportionately old (median age above 25 years), disproportionately concentrated in a handful of operators, and disproportionately scheduled. The Atlantic Council's 2023 review of critical undersea infrastructure flagged the repair-fleet bottleneck as the single most underreported strategic vulnerability in the cable system.³

Seven days is acceptable for a single fault on a route with redundant capacity. Four months is the time it takes for the Red Sea to discover that the global cable repair industry was not engineered for simultaneous multi-fault incidents in a war zone. The shadow-fleet doctrine is, among other things, a low-cost way to learn whether the defender is in the seven-day or the four-month regime.

NATO's response.

NATO formally established the Critical Undersea Infrastructure Cell in February 2024 within Allied Maritime Command, with operational coordination by Northwood headquarters in the United Kingdom. The Cell's remit covers situational awareness, allied vessel coordination during incidents, and information sharing with operators. The Cell does not own cables, does not own repair ships, and does not own enforcement authority outside flag-state and territorial-waters jurisdiction. It is institutional progress; it is not the repair fleet.²

EU member states have accelerated landing-station hardening and data-cable diversification programmes through the European Union Strategic Compass and a number of national initiatives. Spain's response in particular has been to push for additional cable landings on the Atlantic coast, away from the densely contested choke of the Strait of Gibraltar; the Marea cable's Bilbao landing is one example. The pattern across the alliance is recognisable: institutional response in 2024, programme funding in 2025, fielded capability later. The adversary's cycle is faster than that.

The hyperscaler shift.

If the Baltic theatre is the operational lesson of the past 30 months, the hyperscaler ownership shift is the structural one. New trans-oceanic capacity through 2026 is overwhelmingly anchored by Google, Meta, Microsoft and Amazon Web Services. The cables involved (2Africa, Marea, Dunant, Equiano, Anjana and the multiple new Pacific systems) constitute the spine of global hyperscale interconnection. The interesting effect of this shift is that it has, almost incidentally, produced more redundancy in the Atlantic basin than 30 years of state-led capacity planning ever achieved.

This is the small irony at the centre of the libertarian read on cable resilience. The strategic-asset-protection regime that 19th-century treaty law contemplates is not what produced the recent capacity build. The capacity build was produced by hyperscaler economics, by the cost of inter-region replication for cloud services, and by the willingness of Google and Meta to write multi-billion-dollar cheques without waiting for state coordination. Resilience by market diversity arrived because the market wanted it. Resilience by state coordination did not, because the state was waiting for a treaty.

The libertarian read.

The architecture of cable resilience that actually works is redundancy by diversity: more cables, more landing points, more owners, more legal regimes. The architecture that has been proposed in much European discourse is concentration: a single governance authority, a single repair regime, a single financing pool. The first is harder to break. The second is easier to manage. Adversary doctrine targets the manageable surface, not the diverse one.

The hyperscaler ownership shift, with all its implications for sovereignty over a piece of EU-landed infrastructure, has had the practical effect of adding diversity faster than any state-coordinated alternative would have. The Chronicle's view is that this is, on balance, the right outcome for resilience and the wrong outcome for European sovereignty, and that the correct response is for European telcos and consortia to match the hyperscaler build rate rather than to argue for slowing it. The first option is hard. The second is impossible. Choosing on the wrong side of that distinction has been a recurring habit of European technology policy, and the cable industry is the next chapter in which it will be made.

What it means for builders.

If you are designing systems that depend on global connectivity, the doctrinal lessons of the cable infrastructure are the following.

Assume connectivity is non-uniform across regions. Inter-region replication that ships through a single trans-oceanic cable is not redundant; it is a single fault path with a different label. Trace your real bandwidth path, not the abstract one your provider's marketing diagram shows.

Latency is the cable layer's verdict. Software architectures that assume sub-100ms inter-region round-trips are betting on the cable layer. Those bets are usually safe; they are not always safe. Architectures that degrade gracefully under 500ms are robust against minor incidents and acceptable to most users.

Repair time is your worst-case planning horizon. If your business interruption tolerance is one week and the regional repair-time worst case is four months, you are exposed by a factor of seventeen. Either the tolerance has to relax or the redundancy has to come from a different geography.

Hyperscaler diversification is real diversification. A workload that runs on AWS in Ireland and on Google Cloud in Belgium, with cross-region failover, has approximately twice the cable-layer fault tolerance of a workload that runs on a single hyperscaler in two regions of the same continent. The premium is small. The optionality is large.

Closer.

The cable layer is the strategic infrastructure most users have never thought about. The shadow-fleet doctrine has begun to make the layer visible, expensively. The Chronicle's view is that the visibility is overdue and the architectural response is owed. The libertarian shape of that response (more diversity, faster build, less treaty) is the shape the past 30 months of incidents have implicitly endorsed. Whether European policy is willing to read the endorsement is the next chapter.

Tobias Hager

Founder · CEO · Sole engineer · mi media intelligence SL

Tobias designs and operates every system inside the house. He builds AI-native products and the offensive/defensive cyber capability stack the company runs alongside its research thread toward agentic generative reasoning. He writes the Chronicle from inside the operational stack rather than from outside of it.

tobias-hager.com LinkedIn contact@mi.network

mi media intelligence SL · Palma de Mallorca · April 2026 · SIGIL CHR-0004