The physical information layer — undersea cables & the space/ISR layer

Web-verified 2026-06-11. Structured + sources: geopolitics-cables-space-layer.json. Overlay — evidence-graded, excluded from the proofs. Extends spec-telecom-satellite; connects to geopolitics-allied-intel-nodes (Singapore/HK cables), geopolitics-russia-energy-arctic + geopolitics-contested-resource-states (shadow fleet), spec-disclosures-surveillance.

The data chokepoint, made physical and kinetic. ~99% of intercontinental data rides undersea cables that are being cut with plausible deniability; and the orbital ISR/comms layer is concentrating into one private actor. Both are contested, under-defended chokepoints on the information substrate the AI/intel economy assumes is a given.

Undersea cables — cheap to cut, slow to fix

• The cuts. ~10 Baltic cables cut since 2022, 7 of them Nov-2024→Jan-2025 (e.g., BCS East-West Sweden–Lithuania and C-Lion1 Finland–Germany within 24h; the Chinese vessel Yi Peng 3, departing Russia's Ust-Luga, a prime suspect). Around Taiwan: 5 incidents in 2024–25; the coast guard blacklists 96 China-linked vessels (flags of convenience) across 24 cables. ~6 Sep 2025: 15 cables severed at Bab el-Mandeb (Red Sea) → Middle East / South Asia outages (fact).
• Deniability is the weapon. Anchor-dragging is low-sophistication and deniable; US/EU officials have called some Baltic damage accidental rather than sabotage — the ambiguity is the point (fact + contested).
• Purpose-built capability. Beyond improvised dragging, China is developing purpose-built cable-cutting capability (a patented compact deep-sea cutter; CSIS flagged a new cutting ship) — moving from gray-zone to wartime capability.
• The real chokepoint is repair. The global cable-repair fleet is aging (~¼ of ships >40 years old) and too small (most ships lay new cable, not repair); Asia-Pacific repairs take ~30 days vs ~15 in North America (permitting). Cutting is cheap; fixing is slow and capacity-constrained. NATO's Baltic Sentry (14 Jan 2025) — frigates, patrol aircraft, naval drones — is the deterrent response.

The space / ISR layer — concentration into one actor

SpaceX Starshield (a Starlink-derived military bus) builds the NRO's proliferated constellation — target tracking, optical/radio reconnaissance, missile warning. ≥183 Starshield satellites launched by 2025; a $1.8B classified contract (2021, revealed 2023) for hundreds of spy satellites; operational from May 2024; launches planned through 2029 (fact). The significance: US military space ISR + comms is concentrating into one private actor (SpaceX) — the same single-point pattern as TSMC (compute) and ASML (litho), now for orbital intelligence and the launch cadence that sustains it — a firm whose owner also sits across the AI core (xAI/Colossus), launch, and satellite comms. (An Oct 2025 NPR report found Starshield satellites emitting an unexplained downlink on uplink frequencies, counter to ITU norms — surfaced by amateur astronomers; capability/oversight opacity, contested.)

The pattern

The information substrate the AI/intel economy treats as a given — cables carrying ~99% of intercontinental data, plus the orbital ISR/comms layer — is in fact a set of contested, under-defended, increasingly concentrated chokepoints. Cables: cheap deniable sabotage (shadow-fleet / anchor-drag) against a too-small, aging repair fleet. Space: ISR + comms funneling into a single private actor. Same lesson as Taiwan / ASML / HBM / HALEU, applied to the data layer — the AI build silently assumes the bits keep flowing.

← Research index · structured data: geopolitics-cables-space-layer.json · geopolitics-cables-space-layer.md