Every time SpaceX launches another batch of Starlink satellites, the headlines celebrate the expanding constellation. Amazon Leo races toward its FCC-mandated 1,618-satellite milestone by July 2026. OneWeb touts global coverage. But behind every LEO mega-constellation is a quieter, more prosaic constraint: the ground gateway network.

Gateway stations—the facilities that connect orbiting satellites to terrestrial fiber infrastructure—are the indispensable link between space-based capacity and the actual internet. Without them, a constellation is just an expensive array of orbiting radios with nowhere to route the traffic. Yet building this ground segment at the scale required to support tens of thousands of satellites is proving harder than the satellite deployments themselves.

The Fiber Requirement Nobody Talks About

A gateway station is deceptively simple in concept. Large antennas point skyward, communicating with passing satellites over feeder links in the Ka or Q/V bands. The data flows down from orbit, gets handed off to high-capacity fiber optic lines, and enters the internet backbone. But that terrestrial fiber connection is not optional—it is the gateway’s entire purpose. A LEO satellite operator must either deploy gateways where robust fiber already exists, or build that fiber infrastructure themselves.

Starlink operates approximately 150 gateways worldwide as of mid-2026, with another 13 under construction and 19 pending regulatory approval. In the United States alone, the network includes more than 100 sites with over 1,500 antennas. Each location requires negotiation with landowners, construction permits, spectrum licenses, and—crucially—access to high-speed fiber. The closer a gateway sits to end users, the lower the latency and the higher the effective throughput, because data spends less time bouncing between satellites or traversing long terrestrial routes. This proximity requirement means gateways cannot be clustered in a few convenient locations; they must be geographically distributed, multiplying the logistical and regulatory burden.

Amazon Leo faces the same calculus. The company has filed applications for gateway licenses in the United Kingdom (Bude, Cornwall), selected Kenya for its first African gateway, and is building others across multiple continents. Each filing triggers regulatory review, coordination with existing satellite operators, and interference analysis to prove coexistence with co-frequency systems. A single gateway approval can take months or years depending on the jurisdiction, and there is no guarantee of success.

Spectrum: The Ka-Band Traffic Jam

The electromagnetic spectrum allocated to satellite feeder links is finite, and it is getting crowded. Most LEO operators rely on Ka-band frequencies (26.5–40 GHz) for their gateway-to-satellite uplinks and downlinks. At these frequencies, atmospheric attenuation—especially rain fade—is a persistent problem. A heavy rainstorm can dramatically degrade link performance, forcing systems to either over-provision capacity or deploy redundant gateways in different weather zones.

More troubling is the spectrum congestion itself. A February 2026 industry panel warned that the assumption of easily securing Ku or Ka-band rights is obsolete: operators entering the market now are “already too late.” The spectrum, once a vast and forgiving resource, has become a contested industrial corridor. This has accelerated the migration toward Q/V band (40–75 GHz), where more spectrum is available. Ofcom, the UK regulator, opened access to up to 10 GHz of Q/V spectrum for satellite gateways in low-density areas covering 94 percent of the UK landmass. But Q/V brings its own challenges: even greater susceptibility to rain fade, more expensive ground equipment, and less mature technology.

The technical response has been multi-antenna gateway stations, which use multiple large antennas at a single site to manage feeder links with many satellites simultaneously. This increases capacity but also increases cost and complexity. Every additional antenna requires more real estate, more power, more cooling, and more fiber backhaul capacity. The gateway becomes a data center in its own right.

The Economics of Global Coverage

Techno-economic models of LEO constellations consistently highlight a sobering reality: these systems are most competitive in low-density areas, typically below 0.1 users per square kilometer. In denser regions, terrestrial fiber is almost always cheaper per bit delivered. This creates a strategic tension. To serve rural and remote users profitably, operators need global gateway coverage—but deploying gateways in low-density markets means building expensive infrastructure in places with limited revenue potential.

The paradox is that gateways themselves require the very thing LEO satellites are meant to bypass: terrestrial infrastructure. A gateway in rural Montana still needs high-speed fiber running to it, which means either paying an incumbent carrier for access or trenching fiber across empty land. The capital expenditure can be justified in aggregate across a global network, but it remains a hidden cost that does not scale as neatly as satellite production. Launch costs have plummeted thanks to reusable rockets; gateway deployment costs have not.

Starlink’s advantage here is its vertical integration with SpaceX and its ability to move quickly without external coordination. Amazon Leo benefits from integration with AWS infrastructure and existing data center relationships, potentially easing fiber backhaul arrangements. Smaller operators—OneWeb, Telesat—must negotiate every site, every fiber contract, every spectrum license with less leverage. The gateway bottleneck disadvantages late entrants even more than the satellite manufacturing ramp does.

Regulation: The Long Tail of Approvals

Satellite spectrum coordination is governed by the International Telecommunication Union, but actual gateway licenses are issued by national regulators. An operator seeking truly global service must file applications in dozens of countries, each with its own processes, timelines, and technical requirements. Some approvals arrive in months; others stretch into years. In markets where regulatory capacity is limited or political considerations weigh heavily, gateway approvals can stall indefinitely.

The result is a patchwork. Starlink’s 150 gateways provide robust coverage across North America, Europe, parts of South America, and selected markets in Asia and Africa. But gaps remain, and those gaps constrain where the service can be offered at full performance. Amazon Leo’s selection of Kenya for its first African gateway, announced in June 2026, signals the importance of regulatory friendliness and existing infrastructure. Where fiber is scarce and regulators are slow, even a satellite operator with billions in capital cannot quickly deliver service.

What This Means for the LEO Race

The gateway bottleneck does not prevent LEO constellations from succeeding, but it does shape who can succeed and where. Operators with deep pockets, strong government relationships, and existing terrestrial infrastructure partnerships have a significant edge. Starlink’s early lead in gateway deployment translates into a lead in effective service availability, independent of satellite count. Amazon Leo’s integration with AWS may allow it to co-locate gateways with cloud infrastructure, reducing incremental costs. OneWeb and Telesat must build gateway networks largely from scratch, or partner with local carriers who may also be competitors.

The satellites are the visible part of the constellation, but the ground network determines whether those satellites can actually deliver internet service. Every additional gateway expands coverage, reduces latency, and adds redundancy—but it also requires capital, fiber, spectrum, and regulatory patience. The companies that master this ground game will define the competitive landscape as surely as those that launch the most satellites. In LEO, the hardest infrastructure problem is still firmly on Earth.

Sources