30 July 2025
Recent years have seen increasing interest in low-earth orbit (LEO) satellite constellations. With Starlink’s recent announcement of v3 satellites offering as much as 1Tbps capacity per satellite, it is key to understand how satellite connectivity could disrupt terrestrial networks.
In this article, we review the fixed and mobile capacity unlocked by LEO constellations and assess the potential impact on terrestrial networks.
What is LEO satellite connectivity?
While satellite connectivity is nothing new, with geostationary orbit (GSO) satellites providing communication services since the 1960s, the connectivity offering of LEO satellites has dramatically improved in recent years. SpaceX’s Starlink LEO constellation has over six million subscribers, and partnerships have been established between LEO operators, MNOs, and vendors. High-profile collaborators include AST SpaceMobile and Vodafone, Globalstar and Apple, Starlink and T-Mobile.
GSO and LEO satellites have different characteristics:
- Latency and power: GSO orbits are ~36000km above the Earth’s surface, while the most used LEO orbit is below 600km. Therefore, LEO satellites decrease round-trip latency by ~50× and reduce power loss between satellite and terminal.
- Coverage: GSO satellites provide constant coverage to a large footprint, while LEO satellites have small footprints and orbit many times each day. LEO satellites are deployed in large constellations, with handovers between satellites providing uninterrupted connectivity.
Satellites can provide near-global fixed (or portable) and mobile (direct-to-device, or D2D) connectivity, while their resilience to crises such as earthquakes and wildfires enables emergency communications channels. These benefits raise the prospect of competition with terrestrial networks.
How much capacity can LEO satellites provide?
To explore the impact of satellite connectivity on terrestrial networks, we estimate the capacity of Starlink’s LEO constellation: its fixed (or vehicular mounted mobile) offering utilising 2GHz of Ku-band spectrum and its D2D partnership in the USA with T-Mobile, utilising 2×5MHz of PCS spectrum1.
D2D Capacity
As mobile phones have omni-directional antennas, received D2D signals will be weak, enabling QPSK modulation at best. We would expect 5MHz of downlink spectrum to provide at best 10Mbps of capacity, although the real spectral efficiency is not known. With eight transmission beams, each satellite has an effective capacity of 80Mbps2.
Starlink’s initial request for temporary authorisation is for 840 D2D satellites3, although this is well below the initially planned Starlink overall constellation of 12 000 satellites (> 7 000 are now in service) or the potential future constellation size which could be as large as 42 000 satellites. To achieve near-global coverage, each of the planned 840 D2D satellites must cover ~600 000sqkm. Therefore, each satellite will provide a capacity density of 0.00013Mbps/sq. km, or 130bps/sq. km. If Starlink’s original planned fleet of 12 000 satellites was equipped for D2D – reducing each satellite’s footprint – it would achieve 0.0019 Mbps/sq. km, or 1.9kbps/sq. km. If the more recently discussed final fleet size of 42 000 is reached and fully equipped for D2D, the density would be 0.0065Mbps/sq. km, or 6.6kbps/sq. km. – though further improvements could be made if a greater bandwidth of spectrum can be dedicated to D2D, which will be the case if dedicated licensed mobile satellite service (MSS) spectrum is used for LEO D2D services.
To contextualise this, we compare against typical mobile spectrum carriers.
The capacity density provided by a terrestrial mobile network varies from ~0.6 Mbps/sqkm for a low-band only base station, to 1000 Mbps/sqkm for a mmWave small cell. In dense urban areas, with multiple mid-band carriers, we expect a typical capacity density of 450-600 Mbps/sqkm.
To provide similar capacity density to even a low-band carrier, Starlink would require 2×20MHz of mobile spectrum and almost a million satellites – clear beyond any future constellation. As a result, satellite D2D services (using standard devices and mobile spectrum) will likely be limited to voice, messaging, and low-rate data, providing less than one thousandth of the capacity density of even the most minimally equipped terrestrial mobile network and less than one hundred thousandth of a dense terrestrial mobile network.
LEO D2D connectivity is therefore a highly valuable coverage extension for terrestrial mobile networks, but clearly not a replacement for terrestrial capacity for direct-to-device applications.
Fixed capacity
Directional Starlink terminals have greater antenna gain, potentially allowing average spectral efficiencies as high as 3.5bps/Hz. With eight channels of 250MHz each, and eight transmission beams with three channels per beam, each first-generation satellite was estimated to provide 21Gbps4. Figures provided by SpaceX suggest that V1.5 satellites provide 24Gbps, V2 mini satellites provide 96Gbps and the recently announced, but not yet launched, V3 satellites will provide substantial increase to 1Tbps per satellite. Although we have referred to this capacity as ‘fixed’, it can be used on the move if the Starlink terminal is mounted on a vehicle, boat, plane or train. The important distinction between this capacity and D2D capacity, is that it is access via a dedicated Starlink terminal with an advanced directional and steerable antenna, using Ku-band spectrum, rather than a smartphone omnidirectional antenna using mobile spectrum
For global coverage, each of the ~7000 active satellites5 has a footprint of ~73 000sqkm6. Using a contention ratio of 50:1, Starlink could offer its 100Mbps service to just 10 500 households in each footprint with the original satellite constellation.
SpaceX recently announced v3 satellites, with 1Tbps downlink capacity each7. With increased capacity and a contention ratio of 50:1, Starlink could offer its 100Mbps service to 500 000 households in each footprint. Each satellite in the planned fleet of 12 000 will have a footprint of ~42 500sqkm. In rural areas, this footprint could contain over four million households.
In the figure below, we provide estimates for the household coverage that could be achieved in several deployment scenarios assuming a contention ratio of 50:1 – including the potential extension to 42 000 satellites.
Therefore, the current Starlink could not be the main source of broadband connectivity in any but the most remote areas. However, once v3 satellites are deployed, a constellation of 12 000 satellites could meet a third of broadband demand in highly rural areas (35 households per sq km) and a constellation of 42 000 could meet the entire requirements of such areas.
It should be noted that a 42 000 satellite constellation of v3 satellites would deliver a capacity density approaching that of a terrestrial 5G networks (82 Mbps/sq. km vs. 394 Mbps/sq. km for 3.5GHz 5G mobile in the table above). In contrast, a fibre access network, such as XG-PON serving a low-density suburb of 500 households per square km would deliver a capacity density of around 78,000 Mbps/sq. km – around 1,000 times the capacity density of even a 42 000 satellite constellation.
For fixed services in highly rural areas, a future version of Starlink could potentially provide the entire broadband requirements of the area, although it should be noted that the current generation of Starlink satellites cannot do so. A future version of Starlink would therefore be a viable alterative to 5G FWA in highly rural areas with no fibre, DOCSIS or VDSL network. However, it seems unlikely that any future LEO system could compete with fibre broadband.
How will this impact terrestrial networks?
LEO constellations are likely to serve markets with:
- sparse populations (eventually ~35 households per sq km with future LEO constellations)
- limited access to terrestrial networks.
However, terrestrial networks will remain the default form of connectivity:
- satellites provide limited capacity and can only serve the most rural areas
- capacity in a region cannot be scaled. While LEOs can address locally uneven demand, it’s fundamental that LEO capacity is spread over the entire constellation coverage
- in most cases, terrestrial networks are more cost-efficient.
There is scope for competition at the edge of terrestrial coverage. However, these areas are not a core part of terrestrial network business cases – and the availability of satellite services may reduce the need for expensive rural rollout of terrestrial networks.
At the same time, several issues hold LEO constellations back:
- While 3GPP Release 17 included two 5G NTN bands (L-Band and S-Band), there is not yet a unified approach to spectrum for D2D
- Current ITU regulations restrict non-GSO satellite transmission power – although technical studies reviewing these limits are underway, following WRC-23.
Therefore, while LEO constellations promise valuable rural connectivity and resilient emergency communications, they are likely to complement – rather than challenge – terrestrial networks in the majority of markets.
[1] USA FCC, Public notice, DA 23-338, April 2023
[2] D. Rozenvasser and K. Shulakova, Estimation of the Starlink Global Satellite System Capacity, Proceedings of the International Conference on Applied Innovation in IT, Volume 11, Issue 1, March 2023
[3] SpaceX, Request for Special Temporary Authority to conduct experimental operations, FCC Filing, December 2023
[4] S. Pekhterev, “The Bandwidth of the Starlink constellation”, SatMagazine, November 2021
[5] J. McDowell, “Starlink statistics”, accessed via planet4589.org in March 2025
[6] The realistic footprint is much smaller – with satellite beams focussed on high-demand areas – and coverage is concentrated between 55, rather than being truly global. However, this calculation provides an illustrative guide to the achievable household penetration.
[7] Starlink, “Progress 2024”, accessed via stories.starlink.com, January 2025
