The Satellite Internet Story

In 2024, more than 7,000 active satellites are orbiting the Earth, more than half of them part of SpaceX’s Starlink constellation. The LEO satellite internet industry, which did not exist before 2019, now serves more than 4 million subscribers worldwide. The story of how a private rocket company came to dominate the most ambitious telecommunications project of the decade is, in retrospect, a story of reusable launch, of mass-produced satellites, and of an extraordinary regulatory and engineering gamble.

Before Starlink: the GEO era

Before the LEO revolution, satellite internet was provided by geostationary (GEO) satellites orbiting at 35,786 km above the equator. From that altitude, a satellite appears stationary in the sky, which means that a fixed antenna on the ground can point at it without tracking. The trade-off is the latency: at the speed of light, the round-trip from the ground to a GEO satellite and back is about 600 milliseconds — too slow for video calls, online gaming, or any interactive application. The first commercial satellite internet service, Hughes Network Systems‘ DirecPC, was launched in 1996, and was followed by a wave of similar services from Hughes, Viasat, and SES. The technology was useful for broadcast (TV, weather data) and for one-way data delivery, but the two-way interactive use case was limited by the latency.[1, 2, 9]

The bandwidth of GEO satellite internet was also limited. A typical satellite had a total throughput of a few gigabits per second, shared among all the users in its footprint. The cost of the user terminal was high ($1,000–$5,000 in 2000), the monthly subscription was high (more than $100), and the speed was modest (a few megabits per second per user). The technology was, in practice, a last-resort option for users in remote areas where terrestrial broadband was unavailable. The total number of subscribers worldwide, in 2018, was perhaps 5 million — a tiny fraction of the global broadband market.[1, 2, 9]

The other approach in the pre-Starlink era was medium Earth orbit (MEO), used by O3b Networks (now part of SES) and by the Globalstar and Iridium constellations. MEO satellites orbit at 2,000–35,786 km, with latencies of 50–200 milliseconds — better than GEO, worse than LEO. The Iridium constellation, launched in the late 1990s, was the first large LEO constellation, with 66 satellites providing global voice and low-speed data service. The constellation famously went bankrupt in 1999, was rescued by a US government contract, and was rebuilt after a series of in-orbit collisions with decommissioned Soviet-era satellites in 2009. The lesson — that LEO constellations are technically feasible but financially risky — was not lost on the next generation of operators.[1, 2, 10]

The LEO revolution

The LEO revolution was made possible by three concurrent developments. The first was the dramatic reduction in launch cost, driven by SpaceX’s Falcon 9 rocket and its reusable first stage. The cost of launching a kilogram to LEO fell from roughly $10,000 in 2010 to roughly $1,500 in 2024, a factor-of-7 reduction. The second was the dramatic reduction in satellite cost, driven by the use of commercial-off-the-shelf components, by the development of phased-array antennas and ion propulsion for small satellites, and by the use of mass production. The cost of a Starlink satellite, in 2024, is estimated at less than $500,000, a fraction of the cost of a traditional GEO communications satellite. The third was the availability of a large amount of spectrum in the Ku and Ka bands, freed up by the transition from analogue to digital broadcasting and by the consolidation of the satellite industry.[2, 3, 11]

LEO satellites orbit at 500–1,200 km, a factor of 30 to 70 closer than GEO. The one-way latency to a LEO satellite is 5–25 milliseconds, comparable to terrestrial broadband. The round-trip latency, including the user terminal, the satellite, the ground station, and the back to the user terminal, is typically 25–50 milliseconds, fast enough for video calls and online gaming. The trade-off is that LEO satellites move quickly across the sky, requiring a constellation of many satellites to provide continuous coverage, and a tracking antenna at the user terminal. The solution is a phased-array antenna, in which the beam is steered electronically, with no moving parts. Modern phased arrays are small enough to fit in a flat-panel antenna about the size of a laptop, and can track multiple satellites simultaneously while switching between them in microseconds.[2, 3, 11]

The LEO revolution has been driven by SpaceX, but the technical foundation was laid by two earlier ventures. Teledesic, founded in 1990 by Bill Gates and Craig McCaw, proposed a constellation of 840 LEO satellites for global broadband. The project raised more than $1 billion but was cancelled in 2002 without launching a single operational satellite. SkyBridge, a similar venture by Alcatel in the late 1990s, was also cancelled. The lessons of these earlier failures — that the launch cost must be low, that the satellite cost must be low, that the user terminal must be cheap — were absorbed by the next generation of operators, and are visible in the design of Starlink, OneWeb, and Kuiper.[2, 3, 12]

Starlink: SpaceX’s megaconstellation

Starlink, announced by SpaceX in 2015 and first launched in 2019, is the largest satellite constellation in history. As of late 2024, the company has launched more than 7,000 satellites, with more than 6,000 operational. The constellation is being expanded continuously, with new satellites launched at a rate of roughly 100 per month. The first generation of Starlink satellites (v0.9, v1.0) was about 260 kg; the second generation (v2 Mini) is about 800 kg, with a larger antenna and a more powerful solar array. The planned third generation (v3) will be about 2,000 kg and will be launched by SpaceX’s Starship rocket.[3, 4, 13]

Each Starlink satellite is a flat panel, about 4 metres long, with a single solar array, a krypton-fueled ion thruster for orbit maintenance and deorbiting, and a phased-array antenna for communication with the user terminals and the ground stations. The satellite uses inter-satellite links (laser crosslinks) to communicate with other satellites in the constellation, allowing data to be routed through the constellation without going through a ground station. The inter-satellite links are a key innovation: they reduce the latency, increase the resilience, and allow the constellation to provide coverage over the ocean and in polar regions, where ground stations are sparse.[3, 4, 13]

The Starlink user terminal, nicknamed “Dishy McFlatface”, is a phased-array antenna about 50 cm across, mounted on a small stand. The terminal costs SpaceX about $1,300 to manufacture (down from $3,000 in 2020), and is sold to subscribers for $599. The terminal self-orients, communicates with the nearest satellite, and provides a typical download speed of 50–200 Mbps and a typical upload speed of 10–20 Mbps. The subscription costs $90–$120 per month, depending on the country. The service is now available in more than 100 countries, with more than 4 million subscribers, and is the largest satellite internet service in history. The revenue, in 2024, is estimated at more than $7 billion per year, and the business is profitable.[3, 4, 14]

The competition: OneWeb, Kuiper, Guowang

The other major LEO constellations are catching up, but the gap is large. Eutelsat OneWeb (the result of a 2023 merger between OneWeb and Eutelsat) operates more than 600 satellites, providing global coverage with a focus on enterprise and government customers. The OneWeb satellites do not have inter-satellite links, and so require a ground station within range of each user. The user terminal is larger and more expensive than Starlink’s, and the service is sold business-to-business rather than directly to consumers. The merged company is now the largest GEO-plus-LEO operator in the world, with the existing Eutelsat GEO fleet providing broadcast and backhaul, and the OneWeb LEO constellation providing low-latency enterprise service.[4, 5, 15]

Amazon Kuiper is the most ambitious new entrant. The project, announced in 2019, plans a constellation of 3,200 satellites in 98 orbital planes, providing global broadband with a focus on consumer and enterprise users. The first two prototype satellites, KuiperSat-1 and KuiperSat-2, were launched in late 2023, and the first production satellites were launched in 2024. The user terminal is being designed in three sizes: a small, portable terminal for travel and emergency use; a standard residential terminal; and an enterprise terminal for business and government. Amazon’s competitive advantage is the integration with its existing businesses (AWS, retail, logistics) and the company’s track record in consumer hardware (Kindle, Echo, Fire TV). The challenge is speed: the FCC license requires half the constellation to be operational by 2026, and the other half by 2029.[5, 6, 15]

China has two major LEO constellations under development. The first, Guowang (“national network”), is a state-led project with a target of nearly 13,000 satellites, intended to provide a sovereign Chinese alternative to Starlink. The second, Qianfan (“thousand sails”), is led by the Shanghai-based company SSSPACE and has a target of more than 14,000 satellites. Both projects have launched a small number of test satellites in 2024, and both are pursuing a fast ramp-up. The strategic motivation is partly commercial (the Chinese consumer market is large) and partly geopolitical (the US has restricted Chinese access to satellite broadband, particularly in contested regions). The combined Chinese constellation, if completed, would rival Starlink in scale.[5, 6, 16]

Other LEO constellations, smaller or more specialised, include Telesat Lightspeed (a Canadian LEO constellation focused on enterprise and government, with 198 satellites planned), Globalstar (a US LEO constellation providing low-speed data and SOS service for Apple iPhones), and a wave of Chinese startups. The total number of LEO satellites, by the end of the decade, is expected to be well above 50,000, with most of the growth coming from Starlink, Kuiper, Guowang, and Qianfan.[5, 6, 16]

Spectrum, orbits, and the Kessler problem

The LEO revolution has put new pressure on the orbital environment. The number of satellites in LEO has roughly tripled since 2019, and the density of operational satellites in the most popular orbits (550 km, 1,100–1,200 km) is now high enough to be a serious concern. The Kessler syndrome — a cascade of collisions in which each collision generates debris that causes more collisions — has been a theoretical concern since the 1970s but is now a practical one. The International Space Station has had to perform collision-avoidance manoeuvres more than 30 times since 2020, mostly to avoid debris from earlier collisions. The first major collision in the modern era — the 2009 Iridium-Cosmos collision — generated more than 1,800 pieces of trackable debris, and was a wake-up call for the industry.[6, 7, 17]

The regulatory response has been uneven. The US Federal Communications Commission, in 2022, adopted a rule requiring satellite operators to deorbit their satellites within 5 years of mission end (down from the previous 25-year guideline). The European Space Agency, the UK Space Agency, and others have adopted similar rules. The enforcement, however, is patchy: a number of operators have failed to deorbit their satellites on time, and the tracking of small debris is incomplete. The most ambitious proposal, from the US Commerce Department, is to transfer space-traffic management from the Department of Defense to a civilian agency, but the transition is years away.[6, 7, 17]

The technical response has been more encouraging. Modern satellites are designed to be deorbited at the end of their life, by firing their thrusters to lower the orbit until atmospheric drag brings them down. The Starlink satellites, for example, have a deorbit thrust profile that brings them down within about 5 years of mission end. The next generation of satellites will be designed to be self-removing, with enough propellant to deorbit themselves even if the primary propulsion fails. The long-term solution, however, requires active debris removal — sending a spacecraft to capture and deorbit dead satellites — and a number of startups (including Astroscale and ClearSpace) are now developing the technology. The economics are still uncertain.[6, 7, 18]

The geopolitics of satellite internet

Satellite internet has become a geopolitical story. The most prominent example is the war in Ukraine, where Starlink has provided connectivity to the Ukrainian military and to civilians, and has become a key part of the war’s communications infrastructure. SpaceX has been broadly supportive of the Ukrainian effort, donating thousands of terminals and providing service at no cost, but the company has also occasionally restricted the use of Starlink in specific regions, particularly in Russian-occupied territory. The arrangement has been controversial: some have argued that a private company should not have effective control over the communications infrastructure of a country at war, and that the situation highlights the need for sovereign alternatives.[7, 8, 19]

The response, in Europe, has been a combination of public investment in OneWeb and Kuiper, and a longer-term commitment to a sovereign European LEO constellation, tentatively called IRIS². The response in China has been the Guowang and Qianfan constellations, both of which are intended to be sovereign and to be available globally, particularly in the Global South. The response in Russia has been more limited, with a plan for a small constellation called Sfera that has been slow to materialise. The long-term outcome is a world in which the orbital internet is dominated by two or three large constellations, each tied to a major geopolitical bloc, and in which the use of satellite internet is a matter of national policy.[7, 8, 19]

The next decade of satellite internet will be shaped by three trends. The first is the continued growth of the constellations, with the total number of satellites in LEO expected to be well above 50,000 by 2030. The second is the convergence of satellite and terrestrial networks, with 5G and 6G standards being designed to include satellite access as a native component, and with user terminals that can switch seamlessly between cellular and satellite. The third is the gradual erosion of the distinction between communications and observation, with the major satellite operators planning to add Earth-observation, navigation, and timing services to their offerings. The orbital layer of the internet is, in other words, becoming as important as the terrestrial layer, and the geopolitics of orbit is becoming a central part of the geopolitics of information.[7, 8, 20]

Comparison: leading LEO constellations (2024)

Constellation	Operator	Headquarters	Operational satellites (2024)	Planned total	Approx. subscribers
Starlink	SpaceX	Hawthorne, California	~6,000	~12,000 (eventually 30,000+)	~4 million
Eutelsat OneWeb	Eutelsat Group	Paris, France	~600	~650	B2B (not disclosed)
Kuiper	Amazon	Seattle, Washington	~2 (prototypes)	~3,200	Not yet launched
Guowang	China SatNet (state-led)	Beijing, China	~10 (test)	~13,000	Not yet launched
Qianfan	SSSPACE	Shanghai, China	~18 (test)	~14,000	Not yet launched
Telesat Lightspeed	Telesat	Ottawa, Canada	0 (in development)	~198	B2B (planned)

Key numbers

• ~7,000: active satellites in orbit at end of 2024.
• ~6,000: Starlink satellites operational (2024).
• ~4 million: Starlink subscribers (2024).
• ~$7 billion: estimated Starlink annual revenue (2024).
• ~$1,500: cost to launch a kilogram to LEO on Falcon 9 (2024).
• ~$500,000: estimated cost to manufacture a Starlink satellite (2024).
• ~25–50 ms: round-trip latency for Starlink.
• ~600 ms: round-trip latency for GEO satellite internet (pre-Starlink).

References

1. Wikipedia, “Satellite Internet access.” (Standard reference on satellite internet, with a section on the pre-Starlink GEO era.) https://en.wikipedia.org/wiki/Satellite_Internet_access
2. Wikipedia, “Low Earth orbit.” (Standard reference on LEO and the LEO satellite industry.) https://en.wikipedia.org/wiki/Low_Earth_orbit
3. Wikipedia, “Starlink.” (Standard reference on Starlink, with revenue and subscriber data.) https://en.wikipedia.org/wiki/Starlink
4. Wikipedia, “SpaceX.” (Standard reference on SpaceX and its launch business.) https://en.wikipedia.org/wiki/SpaceX
5. Wikipedia, “Kuiper Systems.” (Standard reference on Amazon Kuiper.) https://en.wikipedia.org/wiki/Kuiper_Systems
6. Wikipedia, “Guowang.” (Standard reference on China’s Guowang LEO constellation.) https://en.wikipedia.org/wiki/Guowang
7. Wikipedia, “Kessler syndrome.” (Standard reference on the Kessler syndrome and orbital debris.) https://en.wikipedia.org/wiki/Kessler_syndrome
8. Wikipedia, “Satellite constellation.” (Standard reference on satellite constellations and the orbital environment.) https://en.wikipedia.org/wiki/Satellite_constellation
9. Wikipedia, “Geostationary orbit.” (Standard reference on the GEO orbit used by traditional satellite internet.) https://en.wikipedia.org/wiki/Geostationary_orbit
10. Wikipedia, “Iridium satellite constellation.” (Standard reference on Iridium, the first large LEO constellation.) https://en.wikipedia.org/wiki/Iridium_satellite_constellation
11. Wikipedia, “Falcon 9.” (Standard reference on the Falcon 9 rocket that enabled the LEO revolution.) https://en.wikipedia.org/wiki/Falcon_9
12. Wikipedia, “Teledesic.” (Standard reference on Teledesic, the first attempt at a global LEO broadband constellation.) https://en.wikipedia.org/wiki/Teledesic
13. Wikipedia, “Phased array.” (Standard reference on phased-array antennas used in modern satellite user terminals.) https://en.wikipedia.org/wiki/Phased_array
14. Wikipedia, “Ion thruster.” (Standard reference on ion thrusters used for satellite orbit maintenance and deorbiting.) https://en.wikipedia.org/wiki/Ion_thruster
15. Wikipedia, “OneWeb.” (Standard reference on OneWeb, now part of Eutelsat Group.) https://en.wikipedia.org/wiki/OneWeb
16. Wikipedia, “Qianfan.” (Standard reference on Qianfan, the Shanghai-led LEO constellation.) https://en.wikipedia.org/wiki/Qianfan
17. Wikipedia, “Space debris.” (Standard reference on space debris, including the 2009 Iridium-Cosmos collision.) https://en.wikipedia.org/wiki/Space_debris
18. Wikipedia, “Astroscale.” (Standard reference on Astroscale, a startup developing active debris removal technology.) https://en.wikipedia.org/wiki/Astroscale
19. Wikipedia, “Starlink in the Russo-Ukrainian War.” (Standard reference on the use of Starlink in the war in Ukraine.) https://en.wikipedia.org/wiki/Starlink_in_the_Russo-Ukrainian_War
20. Wikipedia, “Telesat.” (Standard reference on Telesat and the Lightspeed LEO constellation.) https://en.wikipedia.org/wiki/Telesat