The Submarine Cable Map — 600 Systems, 1.5 Million Kilometers, and Where the Internet Is Fragile
Series: Information Infrastructure — The Physical Internet — Part 3 of 8
The Physical Reality Behind “the Cloud”
The metaphor is wrong. There is no cloud. There is a glass thread, roughly the thickness of a garden hose, wrapped in steel armour and polyethylene, lying on the seabed at depths that reach 8,000 metres. There are 570 such systems in active service as of early 2026 (TeleGeography, High), running 1.5 million kilometres of Fiber Optic Transmission across every ocean basin on Earth. They are the substrate of intercontinental data exchange. They carry, by the most-cited figure available, more than 99 percent of all transoceanic internet traffic — a directional claim originating with a 2013 FCC submission and confirmed in subsequent TeleGeography commentary, but never recomputed at precision (TeleGeography, 2023; Medium as a precise figure, High as an order-of-magnitude statement).
Every video call from São Paulo to Lisbon, every financial transaction routed from Singapore to Frankfurt, every government cable from Canberra to London, traverses this network. Disrupt the right four or five points on the map and you do not slow the internet — you partition it. This is the third installment in the Physical Internet series. It maps the topology, the owners, and the chokepoints.
The Scale of the Network
TeleGeography’s 2025 map edition catalogues 597 cable systems and 1,712 landing stations either in service or under construction (High). The numerator moves constantly: legacy cables from the 1990s reach end-of-life and are decommissioned; consortium and hyperscaler builds light up new routes every quarter. The figure to internalise is the active count of around 570 systems, totalling 1.5 million kilometres of fibre — roughly four times the Earth-Moon distance, laid in straight lines through abyssal plains, across continental shelves, and around political obstacles.
The asymmetry of throughput is more arresting than the total. A single modern cable system — Bifrost, Echo, or Grace Hopper — is provisioned for 240 to 352 Tbps of design capacity. That is more bandwidth on one fibre bundle than the entire global internet consumed in 2010. The network is not merely large; it is dense in narrow corridors and thin elsewhere, and the thinness is where strategic risk concentrates.
The Architecture of Hyperscaler Investment
Until roughly 2016, the submarine cable industry was a consortium business. Telecom carriers — Orange, NTT, Telefónica, China Unicom, Singtel — pooled capital across 10 to 30 partners per system, signed 25-year IRU agreements, and shared landing rights. That model still operates, but it has been overtaken in new-build capacity by four American hyperscalers: Google, Meta, Amazon Web Services, and Microsoft.
The shift is structural. Google now operates wholly private cables — Grace Hopper (New York–Bude–Bilbao, 352 Tbps, in service September 2022), Firmina (Myrtle Beach to Argentina with Brazil and Uruguay branches, 14,517 km, in service December 2023), and Equiano (Lisbon to Melkbosstrand, South Africa, with landings in Togo, Nigeria, Namibia, and St Helena, in service September 2022). Meta holds majority or co-majority stakes in 2Africa (the 45,000 km coastal ring around the African continent, core route complete November 2025), MAREA (Virginia Beach–Bilbao, in service since April 2018), and Bifrost (US West Coast to Singapore via Indonesia, RFS October 2025, with Amazon as co-investor, not Google). Microsoft holds fibre pairs on MAREA and on SEA-ME-WE-6 as a non-landing member.
This matters analytically for three reasons. First, the four hyperscalers are now the largest single block of submarine fibre capacity on Earth, larger than any state telecom or carrier consortium. Second, their cables are optimised for their own intra-data-centre traffic — not for general internet routing — which concentrates a disproportionate share of global traffic on routes that exist because Ashburn, Dublin, Singapore, and São Paulo are data-centre clusters, not because those cities are population centres. Third, the cable-builder market has bifurcated: Western hyperscaler systems are built almost exclusively by SubCom (US), Alcatel Submarine Networks (France), and NEC (Japan); Chinese-financed systems run through HMN Technologies, the renamed Huawei Marine Networks subsidiary now controlled by Hengtong Group. The geopolitical reading of any new build begins with which of those five firms holds the construction contract.
Key Systems Mapped
The trans-Atlantic corridor between the US East Coast and the UK/EU is the highest-traffic single segment of the global cable map. MAREA, with 8 fibre pairs and an upgraded design capacity of 200 Tbps, runs from Virginia Beach to Bilbao. Grace Hopper, Google’s 16-fibre-pair private system at 352 Tbps, lands at Bude in Cornwall and at Bilbao. Both terminate in clusters dense with legacy cables — TAT-14, AC-1, AC-2, Apollo, Yellow — making the Virginia–New York–Bude triangle the densest cable landing complex in the world.
The trans-Pacific routes are dominated by hyperscaler consortia. JUPITER (Japan–Oregon/California with a Philippines branch, in service May 2021, six consortium members: Meta, Amazon, NTT Communications, PCCW Global, PLDT, SoftBank) carries 60 Tbps. Echo (Eureka CA–Singapore via Indonesia, 16,051 km, NEC-built) provides 260 Tbps on the California-to-Guam segment, with Google as primary owner and Meta holding an IRU on the Indonesia branch. Bifrost runs a parallel route on the same corridor, redundantly diversifying the Indonesia–US trans-Pacific path. Apricot, expected RFS 2026–2027 after delays, will link Singapore to Japan via Indonesia, the Philippines, Taiwan, and Guam, with fibre pairs split among Google (4), Meta (4), NTT (2), Chunghwa Telecom (1), and PLDT (1).
The Asia–Europe corridor is the most complex theatre on the map. SEA-ME-WE-6 (France–Singapore via the Mediterranean, Red Sea, and Indian Ocean, RFS Q1 2025, SubCom-built, 12 landing members) is the latest in a 40-year sequence of SEA-ME-WE systems. AAE-1 (Hong Kong/Singapore–France, 25,000 km, upgraded 2022 with Infinera ICE6 to exceed 100 Tbps) links a 17-carrier consortium including China Unicom, Etisalat, Reliance Jio, and Telecom Egypt. PEACE Cable (Pakistan–Egypt–Kenya–France with a Singapore extension, in service 22 December 2022) is the strategically distinct entry: 100 percent owned by a Hengtong Group subsidiary, built by HMN Technologies, and used by Beijing as a flagship Digital Silk Road infrastructure project. PEACE is not a consortium cable. It is a Chinese corporate cable with a Chinese state-policy logic.
Africa’s connectivity rests on four backbone systems. 2Africa, the 45,000 km Meta-led ring, is the dominant new entry. Equiano serves the western African coast on Google’s private network. EASSy (Sudan–South Africa, 10,000 km, in service July 2010, 90 percent African-owned with WIOCC as largest shareholder) and SEACOM (South Africa–Djibouti–France/India, 17,000 km, in service July 2009, majority private African ownership through IPS, Remgro, and Sanlam) serve the eastern coast. The continent moved in fifteen years from being the world’s most cable-poor major landmass to one with redundant capacity on both coasts — though landing-station density and terrestrial backhaul remain bottlenecks.
The Chokepoints — Where Traffic Concentrates
The cable map’s strategic importance is not the global total. It is the small number of corridors where the geography forces concentration. Five matter most.
Suez Canal / Red Sea / Bab el-Mandeb. Approximately 14 cable systems land in Egypt and an equivalent count transits Bab el-Mandeb. The CSIS estimate — “over 90 percent of Europe–Asia communications travel through Egypt” — is directional but consistent with cable-by-cable accounting (CSIS, 2024; High). In February 2024, three systems were disrupted simultaneously in the southern Red Sea: SEACOM/TGN-EA, EIG, and AAE-1. Causation remains contested (anchor drag from a vessel previously struck by Houthi missiles is the operating hypothesis; Medium). The incident demonstrated that a single underwater event can degrade between 25 and 40 percent of Europe–Asia bandwidth for weeks. See Red Sea Cable Cuts 2024 for the incident dossier.
Luzon Strait. The 250-km-wide channel between Taiwan and the Philippines carries the trunk routes connecting mainland China, Hong Kong, Taiwan, Japan, and Korea with Southeast Asia, Australia, and the US West Coast. Open-source estimates as of 2020 counted 18-plus international cables transiting the strait; the current figure is almost certainly higher. In December 2006, a magnitude-7.0 earthquake off Hengchun severed seven cables simultaneously and disrupted internet access across East Asia for weeks. The strait combines high cable density, high seismic risk, and acute geopolitical exposure to a Taiwan Strait contingency. See Taiwan Strait Crisis — Strategic Assessment.
Strait of Malacca. The Singapore–Malaysia–Indonesia corridor is the primary routing channel for SEA-ME-WE systems, APCN, FLAG, and a growing volume of hyperscaler builds. Singapore’s role as the Southeast Asian data-centre hub means traffic concentrates here independently of any geographic forcing function — the city-state is the destination, not just the waypoint.
Strait of Hormuz. Cables serving Gulf states — TEAMS, PEACE branches, AAE-1 — transit Hormuz, the same corridor that carries roughly 20 percent of global crude oil flows. Co-location of energy and information chokepoints creates a doctrinal target set for any actor pursuing Economic Chokepoints — Coercive Statecraft simultaneously across multiple domains.
North Atlantic landing cluster. The Virginia Beach–New York–Bude triangle is the highest-traffic single landing complex on Earth. GCHQ Bude sits on this cluster, a strategic location that long predates the modern era and that explains the persistence of British signals-intelligence prominence despite the post-imperial contraction of UK power elsewhere. See Baltic Sea Cable Incidents 2023-2025 for the parallel European story on the European-Russian fringe.
The Luzon Strait Problem
Of the five chokepoints, Luzon is the one analysts most often underweight. Suez is visible because oil tankers transit it. Hormuz is visible because of decades of Iran-Gulf tension. Luzon is a hydrographic feature most readers cannot place on a map. But the strait does three things simultaneously: it carries the cable trunks linking East Asia to the rest of the world; it sits inside the operational area of any plausible Taiwan Strait contingency; and it is the corridor through which any PLA Navy interdiction effort would have to operate.
A Taiwan crisis would not require a kinetic strike on cables. It would require only the assertion of an exclusion zone around the island. Cable repair ships — there are roughly 60 in service globally, a fleet of which fewer than 20 are available in the Asia-Pacific theatre at any moment — cannot operate in a contested maritime environment. A cut in the Luzon Strait under exclusion-zone conditions would not be repaired in weeks. It would be repaired when the zone lifted.
This is the analytical point. The vulnerability is not the cable. The vulnerability is the repair logistics.
The Egypt Factor
Why does so much Europe–Asia traffic transit a single country? The answer is geographic determinism reinforced by historical contingency. The shortest fibre path from Mumbai to Marseille goes through the Red Sea, the Suez Canal, and the Mediterranean. The Cape of Good Hope alternative adds roughly 8,000 km of cable, which adds latency and capital cost. Overland fibre through Russia exists but is politically constrained, technically constrained by repeater spacing and crossing arrangements, and currently degraded by the Ukraine War.
Egypt has converted geographic position into rent. Telecom Egypt’s WeConnect division charges cable consortia for landing rights, for terrestrial crossing between Mediterranean and Red Sea termini, and for ongoing capacity rights. The fees are non-trivial — public estimates suggest hundreds of millions of dollars per major cable system over the 25-year design life — and the country has used this revenue stream to position itself as an indispensable transit state.
The strategic implication: any disruption to Egyptian state continuity, or any meaningful conflict with the regime in Cairo, immediately becomes an Asia–Europe connectivity crisis. The Egyptian state is a counterparty risk in the architecture of the global internet.
Landing Stations as Strategic Nodes
The 1,712 landing stations are where the cable becomes vulnerable in a different way. At sea, cables are difficult to reach and require specialised vessels to interdict. On land, the cable terminates in a building with a switchroom, power systems, and a backhaul connection to a national carrier. That building is a fixed address. It can be photographed, mapped, surveilled, raided, or bombed.
Three patterns recur in the global landing-station map. First, landing stations cluster: Virginia Beach hosts more than 12 systems; Marseille hosts more than 15; Singapore hosts more than 20. The clustering is driven by submarine geography (continental shelf gradients, undersea ridges) and by terrestrial logic (data-centre proximity). Second, the landing station’s national jurisdiction matters more than the cable consortium’s nominal ownership — a German-owned cable landing in Egypt operates under Egyptian state authority, with all the lawful-intercept and emergency-control implications that follow. Third, hyperscaler landings increasingly co-locate with hyperscaler regional data centres. The cable does not terminate in a carrier-neutral facility and then route to AWS; it terminates inside the AWS regional fabric directly. This is a vertical integration of capacity and compute that did not exist a decade ago.
Strategic Implications
The submarine cable network is the substrate of the global digital economy and the global digital information environment. Its topology is not the result of optimisation against any single objective. It is the product of geography, of consortium economics through the 2010s, and of hyperscaler capital expenditure thereafter. The result is a network with three structural features that any strategic analyst must internalise.
First, the network is concentrated. A small number of corridors carry a disproportionate share of intercontinental traffic. Four or five chokepoints, properly mapped, account for the majority of global resilience risk.
Second, the network is unevenly owned. The hyperscaler share of new-build capacity is now dominant on the trans-Atlantic and trans-Pacific corridors. The Chinese state, operating through Hengtong/HMN Technologies and through PRC carrier participation in consortia, is the second-largest single block. State telecoms in Europe, Japan, and the Gulf hold the balance. Smaller states and most of the Global South are price-takers in a market structured by four American firms and one Chinese conglomerate.
Third, the network’s repair logistics are fragile. The global cable-repair vessel fleet is small, ageing, and concentrated in a handful of operators. In any contested maritime environment, repair becomes infeasible. The cables themselves are resilient to many threats; the operations that maintain them are not.
The next installment in this series maps the cable-laying and cable-repair vessel fleet — the world’s roughly 60 specialised ships, who owns them, where they are based, and what their operational tempo looks like during the cable disruption events of 2023–2025. See SYNTHESIS for the series index and cross-references.
Sources
- TeleGeography, Submarine Cable Map and Submarine Cable 100 data products, 2025 and early 2026 editions. High confidence for system counts, landing-station counts, route geography, and design capacities. telegeography.com
- TeleGeography commentary on the “99% of intercontinental traffic” figure, 2023 (originally FCC 2013 submission). High as order-of-magnitude; Medium as precise figure.
- 2Africa consortium press releases, Meta and partners, 2020–2025. High for route, length, capacity, and partner list.
- Google Cloud Network announcements for Grace Hopper, Firmina, and Equiano, 2020–2023. High.
- Meta Engineering blog posts on MAREA, Bifrost, and 2Africa, 2018–2025. High.
- SubCom and NEC supply-side press releases on Echo, JUPITER, SEA-ME-WE-6, Grace Hopper, Firmina. High for build status.
- HMN Technologies / Hengtong Group corporate disclosures on PEACE Cable, 2022. High for ownership; Medium for capacity (limited public technical disclosure).
- CSIS, Strategic Importance of Subsea Cables, 2024. High for Egypt transit estimate.
- Reports on the February 2024 Red Sea cable disruptions: Reuters, AP, Doreen Bogdan-Martin (ITU) statements, February–March 2024. High for fact of disruption; Medium for causation attribution.
- 2006 Hengchun earthquake cable-impact reports: APT, IEEE retrospectives. High.
- Bifrost consortium composition: TeleGeography database and Meta/Amazon/Telin/Keppel announcements, 2021–2025. High — corrects widespread misattribution to Google.
- JUPITER consortium membership: official RFS announcement May 2021. High.
- Apricot consortium membership: NEC announcement and TeleGeography catalogue. High — corrects KDDI misattribution.
- Equiano route: Google Cloud announcement, 2019 and 2022. High — Portugal to South Africa, no US endpoint.
Series: Information Infrastructure — The Physical Internet Part 1: Series Overview Part 2: Fiber Optic Transmission — How Glass Carries the World’s Data Part 3: The Submarine Cable Map (this article) Part 4: Cable-Laying Vessels — The 60 Ships That Maintain the Internet (forthcoming)