Last updated: July 10, 2026
Introduction
Starlink is easy to picture as an internet system in space: satellites overhead, a flat user terminal on a roof, and broadband arriving from orbit. That picture is partly correct, but incomplete. A satellite internet network is also a ground network. It needs user terminals, gateway earth stations, spectrum licenses, fiber backhaul, power, site diversity, interference coordination, network routing, and regulatory permission in each market.
The hidden infrastructure matters because the satellite is only one relay in the path. A user’s packet may travel from a home terminal to a satellite, then to a gateway, then into terrestrial internet infrastructure. In some cases, inter-satellite optical links can carry traffic across the constellation before it reaches a ground exit point. Even then, the traffic eventually has to enter or leave Earth through an authorized ground system. Space internet is not an escape from ground infrastructure. It is a new arrangement of ground infrastructure and orbital relays.
Spectrum is the other invisible structure. Starlink is not simply “using radio.” It uses different bands for different jobs: user links, gateway links, tracking, telemetry, command, and newer mobile or supplemental coverage services. FCC authorizations for Gen2 Starlink list Ku-, Ka-, V-, E-, and W-band operations, plus mobile satellite and Supplemental Coverage from Space elements in lower cellular-adjacent frequencies. Each band has different propagation, antenna, interference, and capacity tradeoffs.
This article explains why gateways and spectrum are central to Starlink’s realism. The science-fiction image is a sky network. The engineering reality is a network whose performance depends on carefully placed ground apertures and carefully divided radio bands.
Assumptions Used for the Estimates
The estimates use the free-space path loss formula:
FSPL(dB) = 32.44 + 20log10(frequency in MHz) + 20log10(distance in km)
The comparison distance is 500 km. That is a simplified slant range, not a claim that every user or gateway link operates at exactly that distance. A satellite directly overhead at 500 km has a shorter path than a satellite near the edge of a user’s usable sky. At lower elevation angles, the slant range, atmospheric path, and pointing challenge all increase.
The article uses calculation-reference estimates for 12 GHz, 20 GHz, and 80 GHz links at 500 km: about 168.0 dB, 172.4 dB, and 184.5 dB of free-space path loss respectively. These are physical estimates, not mission predictions.
For antenna gain, the article uses a simple circular-aperture estimate:
Gain = 10log10(efficiency x (pi x diameter / wavelength)^2)
This is not a complete Starlink antenna model. Real Starlink user terminals use phased-array behavior, and gateway earth stations have real feed, polarization, tracking, amplifier, thermal, and regulatory constraints. The aperture equation is included only to explain why a 1.85 meter gateway antenna is physically different from a small consumer terminal or a phone.
Facts about authorized frequencies and ground station filings come from FCC documents. The discussion of bottlenecks is an engineering interpretation based on those facts and on basic link budget behavior.
Physical Calculation
Free-space path loss by band
At 500 km, the calculation reference gives the following free-space path loss estimates:
2 GHz: about 152.4 dB
12 GHz: about 168.0 dB
20 GHz: about 172.4 dB
80 GHz: about 184.5 dB
The differences are large. A 12 GHz link has about 15.6 dB more free-space path loss than a 2 GHz link at the same distance. A 20 GHz link has about 20.0 dB more. An 80 GHz link has about 32.1 dB more. Every 3 dB is roughly a factor of two in power, so these differences are not small tuning details.
Why use higher frequencies at all? Because higher bands can offer wider bandwidths and can be paired with high-gain directional antennas. For a fixed physical aperture, antenna gain rises as wavelength gets smaller. That is the gateway trade: the path loss is higher at Ka-band or E-band, but a properly built ground antenna can form a narrow, high-gain beam, and wider channels can carry more aggregate data if the link budget and interference rules allow it.
Why larger gateway apertures matter
Consider a simplified 1.85 meter gateway antenna. FCC public notices list Starlink gateway earth station filings using 1.85 meter SpaceX antennas in Ka-band ranges such as 27.5-29.1 GHz and 29.5-30.0 GHz for Earth-to-space, with 17.8-18.6 GHz and 18.8-19.3 GHz for space-to-Earth. At 28 GHz, the wavelength is about 0.0107 meters. If a circular aperture of 1.85 meters operated with 60 percent efficiency, the gain estimate would be:
10log10(0.6 x (pi x 1.85 / 0.0107)^2)
That is about 52.5 dBi. This is an illustrative aperture estimate, not an official Starlink antenna specification. But the scale is useful. A 50 dB-class antenna gain can offset a large part of the 170 dB-class path loss. It also concentrates the signal into a narrow direction, which helps frequency reuse and interference control.
Now compare that with a smaller user terminal. If a simplified 0.5 meter aperture at 12 GHz had 50 percent efficiency, with wavelength near 0.025 meters, the gain estimate would be about 33 dBi. Again, this is only an aperture scale, not a Starlink terminal specification. But it explains why a roof or portable terminal can close a link that an ordinary phone cannot close in the same way. The terminal has area, power, pointing, and processing that the phone lacks.
The same physics explains why ground gateways do not disappear just because satellites are advanced. A gateway can be placed on land with clear sky view, professional installation, reliable power, fiber connectivity, cooling, monitoring, and multiple antennas. It can transmit upward with high EIRP under license, receive with high gain, and coordinate with known points of communication. A home terminal is smaller and consumer-managed. A phone is far smaller still.
Geometry, latency, and usable footprint
The orbital geometry also matters. At 550 km altitude, the calculation reference gives an ideal vertical one-way light time of about 1.83 milliseconds, and an ideal vertical ground-satellite-ground light time of about 3.67 milliseconds. That sounds extremely low, and the space segment can indeed be short compared with geostationary satellite paths. But real internet latency also includes terminal processing, satellite processing, routing, gateway location, terrestrial backhaul, congestion, and protocol behavior. A nearby satellite does not guarantee a nearby internet exit.
The footprint calculation has a similar lesson. At 550 km, an ideal horizon footprint radius is about 2,557 km, while a practical 25-degree-elevation footprint radius is about 941 km. The wider theoretical footprint is not the same as high-capacity service. Networks prefer higher elevation angles and narrower beams because they improve link margin, reduce atmospheric path length, and simplify interference management. A satellite may be visible over a large region, but capacity must be divided into beams, channels, time slots, and routing paths.
Operational Reality
FCC documents show the system’s layered structure. FCC 22-91, the 2022 Gen2 Starlink order, described communications between Gen2 satellites and customer user terminals in Ku-band, and between satellites and both user terminals and gateway earth stations in Ka-band. It also discussed proposed Ka-band gateway operations in 17.8-18.6 GHz and 18.8-19.3 GHz space-to-Earth, and 27.5-29.1 GHz and 29.5-30.0 GHz Earth-to-space. That is not a vague statement about internet from space. It is a frequency plan tied to real earth stations and interference rules.
FCC DA 26-36 expanded and modified the U.S. Gen2 authorization. Within that authorization, it listed service and gateway links across Ku-, Ka-, V-, E-, and W-bands, allowed communications with duly authorized gateway earth stations and user terminals on authorized frequencies, and authorized a second tranche of 7,500 Gen2 satellites, bringing the Gen2 authorization described in that order to 15,000 satellites. The same order also included conditions, deferred some requests, and maintained the regulatory structure around harmful interference.
Ground station filings make the infrastructure visible. A 2025 FCC public notice for satellite earth station applications lists SpaceX gateway sites such as Sunset, Texas; Anderson, South Carolina; Olympia, Washington; Phoenix, Arizona; Hawthorne, California; Kiowa, Colorado; and others. The entries list 1.85 meter antennas and Ka-band transmit and receive ranges. Some entries request adding multiple identical antennas. This is what the ground side of a satellite internet network looks like: many licensed antennas at specific places, not just user terminals scattered on rooftops.
Gateways are also network bottlenecks in the practical sense. If a region has many active users but limited gateway visibility, limited backhaul, weather impairments, or constrained spectrum reuse, the satellite layer alone cannot solve the problem. Inter-satellite links can help route around geography, oceans, polar regions, and missing nearby gateways. But they shift the exit point; they do not remove the need for exit points. Somewhere, the traffic must meet terrestrial internet infrastructure.
Weather and frequency add another constraint. Ku-band and Ka-band links are more weather-sensitive than lower cellular bands, and E-band or W-band links face even stronger atmospheric and rain attenuation concerns. Operators can manage this with margins, adaptive coding and modulation, power control, gateway diversity, and routing. The operational point is that a gateway network is not a passive background detail. It is an active part of availability.
Spectrum coordination also limits simple scaling. A satellite cannot transmit unlimited power over unlimited bandwidth wherever it wants. FCC orders discuss equivalent power flux density, power flux density, geostationary arc protection, radio astronomy protection, NASA facilities, processing-round sharing, and limits on co-frequency beams. These constraints exist because many systems use neighboring or overlapping spectrum. Starlink’s capacity grows not only by launching satellites, but by using spectrum more efficiently within a negotiated radio environment.
Operationally, users see the result as speed, latency, reliability, or congestion. Engineers see a chain: terminal to satellite, satellite beam to spectrum channel, satellite to gateway or laser relay, gateway to fiber, and then the wider internet. The weakest link in that chain can become the visible customer experience.
Speculative Lens
One cautious way to imagine Starlink is not as satellites replacing Earth, but as terrestrial networks gaining a moving extension above the atmosphere. The extension still needs hardened ground nodes where data can enter and leave. Gateways are not the glamorous part of the system, but without them the constellation becomes a sky full of relays with too few practical exits.
This changes the way “space internet” should be imagined. The futuristic part is not only thousands of moving satellites. It is the coordination among orbital beams, ground apertures, national spectrum rules, fiber routes, and software-defined routing. The hard question is not just whether a satellite can see a village. It is whether the whole path from that village to the internet has enough margin, capacity, and permission to work at that moment.
In that sense, Starlink is less like a single invention and more like a layered infrastructure project. The user notices the dish. The network depends on the map of gateways, the spectrum plan, and the ability to move traffic through a changing orbital mesh.
What This Does Not Prove
This article does not prove a specific Starlink speed, capacity, or latency in any location. The path loss and antenna calculations are simplified estimates. Real performance depends on terminal model, satellite generation, beam loading, weather, elevation angle, gateway availability, optical inter-satellite routing, local licensing, network congestion, and service plan policies.
It also does not prove that every Starlink packet must always pass through the nearest gateway. Modern Starlink satellites may use optical inter-satellite links, and routing can change depending on network conditions. The claim is narrower: even with inter-satellite links, ground gateways and terrestrial interconnection remain essential to the internet service.
The spectrum discussion also should not be read as a complete legal summary. FCC authorizations are detailed, conditional, and change through later grants, waivers, modifications, and national decisions outside the United States. A frequency listed in an FCC order is not a universal worldwide permission slip.
Finally, this does not mean gateways are a weakness. They are an unavoidable part of making satellite broadband real. The better question is not whether ground infrastructure is needed, but how dense, resilient, diverse, and well-connected that ground infrastructure must be for the promised service level.
Conclusion
Starlink’s visible symbol is the satellite or the user terminal, but its performance depends on a less visible system of gateways and spectrum authorizations. At 500 km, free-space path loss reaches about 168 dB at 12 GHz, 172.4 dB at 20 GHz, and 184.5 dB at 80 GHz. Those losses are manageable only because directional antennas, phased arrays, high-gain gateway apertures, coding, power control, and licensed spectrum work together.
Ground gateways matter because they close high-capacity links and connect the orbital network to the terrestrial internet. Spectrum matters because every link must fit inside a regulated radio environment shared with other satellite, terrestrial, scientific, and government systems. The hidden infrastructure behind space internet is not a footnote. It is the reason the service can exist as more than a satellite passing overhead.
Related reading
- Starlink Ground Stations Explained: How Space Internet Reaches the Web
- Starlink Laser Links Explained: How Satellites Talk to Each Other in Space
- Starlink User Terminals and Electronic Satellite Tracking
- How Starlink Handles Network Congestion as More Users Join
- Is Starlink Turning SpaceX Into a Telecom Company?
Sources
FCC 22-91, SpaceX Gen2 NGSO Satellite System Order and Authorization:
Click to access FCC-22-91A1.pdf
FCC DA 26-36, SpaceX Gen2 Starlink upgrade applications partial grant:
Click to access DA-26-36A1.pdf
FCC Public Notice SES-02767, satellite earth station applications accepted for filing:
Click to access DOC-411304A1.pdf
ESA, Low Earth orbit explainer:
https://www.esa.int/ESA_Multimedia/Images/2020/03/Low_Earth_orbit
Starlink Technology:
https://www.starlink.com/technology
Source access date: July 10, 2026.
Leave a Reply