Part 7 of the series Information Infrastructure — The Physical Internet. See SYNTHESIS for the full arc.
1. The Resilience Claim and Its Limits
Every wave of submarine cable sabotage — the Baltic incidents of 2023–2024, the Red Sea cuts of February 2024, the Taiwan–Matsu severances — produces the same rhetorical reflex: that Low Earth Orbit constellations now provide a credible backup, that the internet has finally acquired its long-promised “Plan B.” The claim is not wrong. It is, however, geometrically narrower than the policy discourse suggests.
LEO constellations have genuinely transformed connectivity in three domains: last-mile access in underserved geographies, rapid restoration in disaster zones, and military communications in environments where terrestrial infrastructure has been destroyed or contested. They have not transformed — and on current physics cannot transform — the global backbone, which remains overwhelmingly subsea. (Fact) Submarine cables carry approximately 99% of intercontinental data traffic, a share that has held stable across the LEO build-out from 2019 to 2026.
This article maps the LEO landscape, quantifies the capacity gap, examines the Ukraine precedent for political conditionality on private satellite coverage, and traces the bifurcation of orbital infrastructure into competing US, Chinese, and EU-adjacent stacks — a pattern that mirrors the broader splinternet logic already visible in Who Owns the Cables — Hyperscalers, State Champions, and the Battle for Submarine Infrastructure.
2. Capacity Fundamentals — Why Cable Still Wins
The capacity comparison is the single fact that determines everything downstream.
A modern trans-oceanic submarine cable system based on wavelength-division multiplexing over coherent optical transmission (see Fiber Optic Transmission) delivers backbone throughput that LEO constellations cannot approach within current spectrum and orbital constraints.
- (Fact) MAREA, the Microsoft–Meta cable running from Virginia Beach to Bilbao, operational since 2017, was designed for 200+ Tbps on a single eight-fiber-pair system. See MAREA Cable.
- (Fact) Starlink’s total system throughput, as self-reported by SpaceX, is approximately 450 Tbps aggregated across the entire constellation of roughly 6,700 operational satellites and all ground gateways. This is a gross figure shared across every user and every market.
The implication is structural. One mid-grade trans-Atlantic cable provides nearly half the aggregate capacity of the world’s largest LEO constellation. A second cable — and there are hundreds in the global 600-system census — surpasses it outright. The orbital medium is fundamentally bandwidth-constrained by spectrum allocation and the physics of phased-array antenna gain; the marine medium is constrained mainly by the cost of laying additional fibers in parallel, a cost that hyperscalers have demonstrated they are willing to absorb at scale.
LEO’s genuine advantage lies elsewhere: (Fact) Starlink latency of 20–40 ms, versus 600+ ms for legacy GEO satellite, makes interactive applications viable from orbit for the first time. This unlocks tactical, edge, and remote use cases that no submarine cable can serve directly. But latency does not substitute for capacity. The two are independent variables, and the global backbone is a capacity problem.
(Assessment, High confidence) LEO constellations should be understood as a last-mile and edge-access layer that complements the cable backbone, not as a competing backbone medium. Policy framing that conflates the two — common in resilience discourse since 2022 — overstates LEO’s strategic substitution value.
3. The LEO Constellation Landscape
The operational LEO field as of mid-2026 contains four serious commercial actors and two Chinese state-directed mega-projects.
Starlink (SpaceX, United States). (Fact) Approximately 6,700 operational satellites as of early 2025, with FCC authorization for up to 12,000 and pending applications extending the long-term envelope toward 42,000. Ku-, Ka-, and V-band spectrum allocations. Ground stations in 60+ countries. Commercial consumer service overlaid with Pentagon, US Air Force, and allied NATO defense contracts via the Starshield military variant. Starlink is the only LEO system currently operating at meaningful scale; the others are catching up or specializing. See Starlink and SpaceX.
Amazon Kuiper (Amazon Web Services, United States). (Fact) FCC license for 3,236 satellites. First mass-production satellites launched in 2025; full commercial service targeted for 2026. The strategic differentiator is integration with AWS cloud services — a direct hyperscaler-to-orbit pipeline that mirrors the hyperscaler subsea cable strategy.
Eutelsat OneWeb (United Kingdom / France). (Fact) 648 operational satellites at 550 km. Eutelsat acquired the full OneWeb stake in 2023, completing the European consolidation. Service positioning has shifted away from mass consumer toward government, enterprise, maritime, and aviation — a deliberate choice to avoid head-on competition with Starlink on price.
Telesat Lightspeed (Canada). (Fact) 198 satellites planned, focused on enterprise and government connectivity for Canada and global Tier-1 telcos. Smallest of the serious Western players; positioned as a wholesale-only carrier rather than a retail brand.
Guowang / SatNet (PRC). (Fact) 13,000-satellite mega-constellation under construction. Operator: China SatCom, a subsidiary of China Satellite Network Group Co., Ltd. Ownership traces directly to SASAC — the State-owned Assets Supervision and Administration Commission — meaning full Chinese state control. Military dual-use designation is explicit in PRC regulatory filings.
Qianfan / “Thousand Sails” (Shanghai, PRC). (Fact) Separate constellation of 1,000+ planned satellites operated by Shanghai Spacecom Satellite Technology (SSST), a partially state-owned, partially commercial entity backed by the Shanghai municipal government. 18 test satellites launched in 2023. Functions as a second Chinese orbital track parallel to Guowang.
The Western field is commercial with state-customer overlays. The Chinese field is state with commercial fronting. This asymmetry conditions everything that follows.
4. Guowang and the Orbital Splinternet
Guowang is the orbital expression of the Digital Silk Road. The strategic logic, visible in PRC regulatory filings and the December 2020 ITU spectrum applications, has three pillars.
First, orbital slot pre-emption. ITU spectrum and orbital-shell rights operate on a first-come, first-served basis with use-it-or-lose-it provisions. Starlink’s filings target the most commercially valuable LEO shells between 540 km and 570 km. Guowang’s filings deliberately overlap and adjacent-overlap these shells, ensuring that any future Chinese constellation has reserved spectrum regardless of which constellation flies first. (Assessment, Medium confidence) Guowang’s primary near-term mission is administrative — to prevent a Western LEO monopoly on critical spectrum — not commercial revenue.
Second, sovereign LEO coverage for Belt and Road Initiative geographies. A state-owned Chinese constellation can offer connectivity to partner governments without routing through US-controlled gateways. This matters strategically because every Starlink terminal in operation is, in principle, subject to US export-control and sanctions enforcement; Guowang offers an alternative jurisdiction. See Guowang and Digital Sovereignty.
Third, military dual-use. The People’s Liberation Army’s Strategic Support Force absorbed satellite communications, ISR, and electronic warfare into a single command in 2015 and reorganized again in 2024. Guowang’s dual-use classification means PLA access to constellation capacity is structural, not negotiated. This is the orbital equivalent of the PRC’s domestic Civil-Military Fusion doctrine applied to space.
(Assessment, Medium confidence) The result is an emerging orbital splinternet: a US/allied commercial stack (Starlink, Kuiper, OneWeb), a Chinese state stack (Guowang, Qianfan), and a smaller EU-adjacent track (Eutelsat). The bifurcation runs along the same fault lines as the submarine cable bifurcation analyzed in Who Owns the Cables — Hyperscalers, State Champions, and the Battle for Submarine Infrastructure. Countries will increasingly be forced to choose a stack, and stack-mixing will become politically constrained.
5. Ukraine — What Battlefield Satellite Dependency Revealed
The Ukraine case is the most thoroughly documented real-world test of LEO connectivity under combat conditions, and it produced two lessons that will shape policy for a decade.
(Fact) SpaceX began deploying Starlink terminals to Ukraine on 26 February 2022, two days after the full-scale Russian invasion, following a public request from Ukrainian Vice Prime Minister Mykhailo Fedorov and coordinated through USAID logistics. By October 2022, approximately 20,000 terminals were operational across Ukraine, according to Fedorov. Use cases included artillery targeting integration, drone command-and-control links, frontline field communications, and civilian internet in regions where terrestrial infrastructure had been destroyed by Russian strikes. See Ukraine War.
The first lesson — Starlink’s tactical utility — is unambiguous. The constellation provided coverage Russia could not effectively jam at scale, restored communications in liberated territories within hours of Ukrainian advances, and offered a resilient command pipeline that survived persistent strikes on terrestrial relay infrastructure.
The second lesson is the one that should reshape doctrine.
(Assessment, High confidence) In September 2023, Elon Musk publicly acknowledged that he had refused a Ukrainian government request to extend Starlink coverage over Crimean territorial waters during a planned Ukrainian naval-drone attack on the Russian Black Sea Fleet at Sevastopol. Musk’s stated reason was a desire to prevent potential Russian nuclear escalation. The disclosure appeared in Walter Isaacson’s biography Elon Musk (2023) and was subsequently confirmed by Musk on X. The operation was reportedly degraded as a result.
The implications are doctrinal. A private operator unilaterally vetoed a sovereign military operation by a state under armed invasion. No US government authorization, statutory authority, or treaty framework was involved in the decision. The veto was based on the personal risk assessment of a single individual.
The Pentagon’s response was to negotiate separate USAF and DoD contracts in 2023–2024 covering Starshield, a militarized Starlink variant with command-and-control flow routed through US government channels rather than SpaceX corporate authorization. (Assessment, Medium confidence) This partially decoupled US-government-supplied terminals from Musk’s personal veto chain, but it did not eliminate the structural dependency: SpaceX remains the constellation operator, and orbital coverage decisions ultimately reside with the operator.
(Assessment, High confidence) The Ukraine precedent has been internalized by every defense ministry observing the war. Sovereign or alliance-controlled LEO capacity has moved from a strategic luxury to a doctrinal requirement. This is driving European interest in IRIS² (the EU’s planned 290-satellite sovereign constellation), French support for Eutelsat OneWeb, and accelerated PRC investment in Guowang.
6. Ground Stations, Spectrum Wars, and Kessler Syndrome
Three hidden vulnerabilities qualify the resilience narrative further.
Ground station dependency. LEO constellations are not free-floating networks. Every satellite must downlink to a terrestrial gateway to interconnect with the global internet. Starlink operates from approximately 100+ ground station nodes worldwide. Inter-satellite laser links (now operational across the Starlink fleet) reduce per-hop ground dependence but do not eliminate the gateway requirement at network egress. (Assessment, High confidence) A coordinated kinetic or cyber strike on key gateway earth stations would degrade constellation service in the affected region within minutes. Constellation diversity does not equal network independence when ground infrastructure remains sovereign-bounded. The same logic that applies to submarine cable landing points — analyzed in Cable Sabotage as Hybrid Warfare — applies to LEO ground gateways.
Spectrum congestion. (Fact) ITU filings show overlapping spectrum claims among SpaceX, OneWeb, Guowang, Kuiper, and several second-tier operators. Ku-, Ka-, and V-band allocations are already contested. As constellations scale toward their licensed satellite counts, in-orbit interference between constellations becomes a real engineering problem rather than a regulatory abstraction. (Gap) No multilateral spectrum coordination regime currently exists for LEO at the scale required; the ITU framework was designed for a small number of GEO operators and is being stress-tested beyond its design assumptions.
Kessler syndrome. (Fact) The November 2021 Russian Cosmos 1408 anti-satellite test generated approximately 1,500 trackable debris fragments — and many more sub-trackable fragments — in low orbit, forcing the International Space Station crew to shelter in return capsules. Kessler syndrome is the cascade scenario in which collision debris generates further collisions, rendering specific orbital shells unusable for decades. (Assessment, Medium confidence) Mega-constellations of 10,000+ satellites per operator dramatically raise the collision-probability surface. A single deliberate ASAT event in a populated LEO shell, or a single uncontrolled cascade, could compromise the entire orbital backbone class. This is an existential risk to the sector and is currently mitigated only by voluntary operator coordination.
7. Arctic Cables and Polar Connectivity — The Next Frontier
The Arctic is the cable industry’s next contested theater, and the dynamics are inverted from the LEO debate: here, the cable is the disruptor and the existing satellite-dependent polar communications are the incumbent.
Far North Fiber / Arctic Connect. (Fact) A Canadian–Japanese–Finnish consortium is planning a polar cable connecting Japan, Alaska, the Canadian Arctic archipelago, Greenland, and Scandinavia. Sea-ice retreat is opening route windows that were physically inaccessible to cable laying until the early 2020s. The cable is not yet operational as of May 2026; permitting, environmental review, and Indigenous-rights consultations in the Canadian Arctic are the principal pacing constraints. Once operational, it will offer a low-latency Asia–Europe route bypassing the Suez–Mediterranean and Strait of Malacca chokepoints entirely.
Russia and the Northern Sea Route. (Fact) Russia has argued in IMO and bilateral fora for leverage over Arctic cable routing through its claimed extended continental shelf and Exclusive Economic Zone. Actual cable landings on Russian Arctic coast remain minimal. (Assessment, Medium confidence) The Russian position is forward-looking: it seeks to establish precedent for transit rights and landing controls before significant infrastructure exists, anticipating a 2030s build-out.
Svalbard. (Fact) Existing undersea fiber connects Svalbard to mainland Norway. The link is critical because Svalbard hosts polar-orbit satellite ground stations operated by KSAT and used by US, European, and commercial Earth-observation constellations — the high latitude provides multiple daily passes for every sun-synchronous satellite. Svalbard’s mixed-sovereignty status under the 1920 Spitsbergen Treaty creates a unique legal environment, and Norwegian authorities have tightened security around the cable following the January 2022 cable cut incident — still officially unattributed but widely assessed as a probable sabotage event in the same period as the Baltic and Shetland incidents.
(Assessment, Medium confidence) Whoever secures the dominant polar cable landing rights — likely some combination of Norway, Canada, Greenland/Denmark, and the United States — will control the northern internet corridor as Arctic ice retreat opens new permanent routes through the 2030s. The Arctic is becoming a new chokepoint geography of the same class as the Mediterranean–Red Sea–Indian Ocean axis.
8. Strategic Implications
Five conclusions follow from the analysis.
The resilience narrative is half-correct. LEO constellations genuinely provide last-mile resilience, disaster recovery, and remote-access capability that no cable can match. They do not and cannot replace the backbone. Policy framing that treats LEO as cable substitution is technically illiterate and should be corrected wherever it appears in official resilience strategies.
Private operator chokepoints are a new doctrinal category. The Ukraine–Crimea precedent demonstrated that a single private executive can veto sovereign military operations of an invaded state. No prior infrastructure category — not cables, not GEO satellites, not internet exchanges — concentrated decision authority this narrowly. Defense ministries must now treat private LEO operators as actors with veto power, not as service providers.
Orbital infrastructure is bifurcating along the same lines as terrestrial. The US/allied commercial stack, the Chinese state stack, and the EU-adjacent track are forming distinct orbital ecosystems with limited interoperability. The same splinternet pressure that has fragmented routing, content, and cable ownership now applies to LEO. See Digital Sovereignty.
Ground station geography is the next vulnerability map. As inter-satellite links proliferate, ground gateway concentration becomes the constellations’ analog to cable landing stations. The next generation of resilience analysis must catalog gateway locations and sovereign jurisdictions, not just satellite counts.
The Arctic is the next contested cable theater. Polar cable routing will reshape Asia–Europe latency and bypass the most-contested chokepoints. The geopolitics will mirror — and partly overlap with — the great-power competition for Northern Sea Route maritime access. This shifts the cable-sabotage threat surface analyzed in Cable Sabotage as Hybrid Warfare northward through the 2030s.
The next installment in this series — the SYNTHESIS — integrates these findings with the preceding articles on fiber transmission, cable geography, ownership, sabotage, regulation, and chokepoints into a unified analytical frame.
9. Sources
Capacity and constellation facts (High confidence):
- SpaceX public filings, FCC IBFS database (Starlink licensing, satellite counts).
- Microsoft and Telxius technical briefings on MAREA, 2017 commissioning documentation.
- TeleGeography Submarine Cable Map and Global Bandwidth Forecast (annual editions through 2025).
- Amazon FCC license filings for Project Kuiper.
- Eutelsat–OneWeb merger disclosures (2023) and Eutelsat investor briefings.
- ITU Radio Regulations Board filings for Guowang/SatNet and Qianfan/SSST constellations (2020–2024).
Ukraine case (High confidence):
- Walter Isaacson, Elon Musk (Simon & Schuster, 2023), chapter on the Crimea/Sevastopol incident.
- Public statements by Mykhailo Fedorov via X and official Ukrainian Ministry of Digital Transformation channels, 2022–2023.
- US Department of Defense Starshield contract disclosures (2023–2024).
- Reuters and Washington Post reporting on Pentagon–SpaceX contractual negotiations.
Kessler / debris (High confidence):
- US Space Command tracking data on Cosmos 1408 ASAT debris.
- NASA Orbital Debris Quarterly News, post-November 2021 editions.
Arctic cables (Medium confidence, evolving):
- Far North Fiber consortium public disclosures (Cinia, Arteria Networks, Far North Digital).
- Norwegian Communications Authority (Nkom) reporting on the January 2022 Svalbard cable incident.
- KSAT public materials on Svalbard ground station operations.
Assessments tagged Medium confidence reflect synthesis across the above primary sources combined with industry analyst reporting; they are not direct quotations from primary documents and should be treated as analytical inferences subject to revision as new evidence emerges.
Gap: No multilateral primary source publicly catalogs all LEO gateway earth-station locations with sovereign-jurisdiction tagging. This is a known collection gap; future open-source efforts (academic and journalist-led) are beginning to address it but coverage is incomplete as of May 2026.