Starlink and inequality

Who could say no to a project that promises internet access for everyone? Starlink, a global Low Earth Orbit (LEO) satellite constellation, has dominated the press this year with announcements of their service availability in a range of African countries. The news has largely been jubilant. Finally, broadband for all. It sounds like the perfect case of a technology solution nailing a previously unsolvable challenge. But all may not be as it seems. In this article, I examine retail (direct to consumer) LEO satellite constellations, such as Starlink and Project Kuiper, and ask the question whether we might be better off without them.

For those of us with access to the internet, it is a marvel. In spite of the trolling, doom-scrolling, fake news, and social media self-absorption, it is still a magnificent tool for human communication and collaboration. From reading research papers to repairing tools to talking with colleagues across the world, it is all accessible in seconds via the internet. It’s as if internet access gives us superpowers. We can find our way in strange cities, watch the surf roll in on a distant shore, transport ourselves into a virtual classroom. And as resources and services grow on the internet, so do our super powers. 

The unfortunate corollary is that those without affordable internet access fall further and further behind. They lose out on economic opportunities, on social connections, and, increasingly, on access to critical government services. As researcher Kentaro Toyama puts it in The Internet and Inequality:

"Technology is a tool; it amplifies existing human capacities. This means that if anything, indiscriminate dissemination of digital technology tends to aggravate inequalities. Technology helps only when there is firm intention economically, politically, culturally to push against the gradient of inequality."

This is a powerful insight. If our technology strategies and policies don’t directly work to diffuse and appropriately distribute the benefits of this magnifying potential, then you get what we have today: a massive concentration of economic and technological power. Perhaps the most important conclusion from this is that simple market forces are not going to provide affordable access to internet for everyone. In fact, market forces alone are likely to continue amplifying the digital divide. 

Starlink is a US company owned by Elon Musk that aims to be perhaps the largest internet service provider in the world by building a global constellation of thousands of satellites that promise to bring broadband to all no matter where you live on the planet. This has been tried before. Bill Gates famously lost his shirt betting on a broadband constellation called Teledesic in the 1990s. It is no small technological and financial feat to launch a coordinated constellation of satellites that delivers ubiquitous broadband. But against all odds, Starlink has done it. They now have over 5000 satellites in orbit and claim to have 2 million active subscribers. When Starlink first became available in early 2021, I placed an order and ran it for nearly a year. I had to see for myself if it was the real deal. Despite a few hiccups, it was impressive: download speeds well over 100 Mbps and low-latency connections meant that videoconferencing, movie streaming, and real-time collaboration were all feasible. Even more impressive was the ease of setup. Just plug it in and point it (in an unencumbered space) at the sky.

Around the world, as Starlink became available, there were happy noises of successful connections being made. Affordable access in sparsely populated, low-income regions of the world is something I have worked on for decades. When people asked my opinion of Starlink, how could I not encourage them to embrace the technology if it was available and affordable? And they have. Political leaders have been courted by Starlink and have enthusiastically embraced the technology. The president of Kenya, extolled its transformative power on social media after a visit to Starlink’s facilities in the US. The Nigerian minister of communications proudly declared that Nigeria “now has 100% broadband coverage” after allowing Starlink a license.

On the African continent, the one notable holdout has been the South African government, that is insisting that Starlink should have at least 30% local ownership by blacks, women, and people living with disabilities. This policy applies to any telecommunications company that would like a license to operate in South Africa. The opposition party and tech community in South Africa have thrown up their arms at the government’s decision to stand in the way of progress. Kenya had a similar requirement but was sweet-talked by Starlink into waiving it.

But What’s a LEO Constellation?

Before going further, some context might be helpful. Historically communication satellites have been launched one at a time and positioned so that they remain above a fixed point on the earth operating further out in space. These are called geosynchronous (GEO) satellites. More recently a new category of satellites has emerged, non-geosynchronous (NGSO) satellites that operate in constellations closer to the earth. Because they are closer to earth, they have to fly at a faster pace to maintain their orbit. As a consequence, many satellites (a constellation) are required to provide continuous access over any given spot on the Earth. NGSO satellites are further broken into the general categories of Middle Earth Orbit (MEO) and Low Earth Orbit (LEO) satellites. Starlink and Project Kuiper fall into this last category operating at an altitude of 500-1200 km as compared with GEO satellites that operate at 35,000 km. The proximity of LEO satellites to the earth means that they have none of the delays that we associate with traditional satellite communication. Where GEO satellites may have delays in communication of a half a second or more, LEO satellite communication delays can be as little as 30ms.

Is There a Problem?

I believe in the importance of satellite technologies as an essential affordable access mechanism, particularly for remote regions. However, not all satellite technologies are created equal and I think we should choose satellite technologies that empower people in the countries where they operate, that help to build complementary economic and technological ecosystems, and that increase the overall autonomy of people and nations. In this article, I am suggesting that retail LEO satellite constellations do comparatively little of this and are likely to extract more value than they contribute in the longer term.

And here I am making a distinction between wholesale LEO initiatives like OneWeb, Telesat Lightspeed, the EU’s planned IRIS2 constellation and others that will offer wholesale access services versus Starlink and Project Kuiper that have been designed for direct-to-consumer retail services. I think wholesale satellite services are more likely to foster local economic ecosystems. They also have the virtue of being orders of magnitude smaller constellations, consisting of hundreds of satellites, which hopefully makes the issue of constellation coexistence less of a challenge.

And let’s not forget the humble GEO stationary satellite. While LEO constellations have been taking the limelight, GEO satellites have been steadily upgraded over the last 5-10 years with new-generation HTS (High Throughput Satellites) versions that are capable of delivering broadband speeds to clients at affordable prices. There is a penalty to pay in latency but for many applications this is not such a big deal. This is happening but it isn’t making the same news as Starlink. In the last month alone, Angola announced that Angosat-2 is connecting 150 locations around the country; Eutelsat Konnect began delivering services in rural Nigeria; and Yahclick commenced broadband services in South Sudan. 

Many concerns have already been raised about Starlink and the sheer volume of satellites that make up the retail LEO constellation. Astronomers are concerned about the extent to which Starlink’s constellation hinders terrestrial astronomical observations, not just visually blocking the night sky but also causing electromagnetic interference with radio astronomy. Others are more concerned by the danger of what would happen if one strayed into the path of another. LEO satellites travel at around 25,000 km/h. For reference, a rifle bullet travels at less than a tenth of that speed. The impact of two satellites colliding could produce the supersonic equivalent of a cluster bomb in space, sending shrapnel in every direction, with each collision creating a new detonation. The impact of this would be to create a debris cloud of potentially millions of pieces of satellite wreckage, effectively rendering near space unusable. This phenomenon is called Kessler Syndrome. While this hasn’t yet happened, there have been near misses and the chance of collision increases as more and more satellites are launched into similar orbits. There are also environmental concerns as to the effect of hundreds, if not thousands, of launches as rocket propellant exhaust interacts with the atmosphere. Any one of these issues might give one enough pause to wonder whether constellations of thousands of satellites are a clever idea. Much has been written about the above. In this article, I want to focus on a few things that I don’t think have received enough attention.

LEO Satellite Constellation Number of proposed satellites Type of Constellation
Starlink 12,000 Retail
Project Kuiper 3,200 Retail
OneWeb 588 Wholesale
Telesat Lightspeed 188 Wholesale
Iris 170 Wholesale
Does Starlink’s Business Model Make Sense?

One of the most appealing features of Starlink is its enduser price. With costs varying substantially from country to country, Starlink monthly subscription fees range between US$25 and US$100; hardware costs range between US$200 and US$650 for the Starlink terminal. This puts Starlink in the same ballpark as the price of monthly residential internet service fees in North America, although with higher equipment costs. For the Global Majority, however, these fees still represent costs that are out of reach for all but a narrow segment of the population.

But do Starlink’s prices reflect their actual costs? We don’t know. For a long time it was natural to assume that what a business charges for a service reflects the provider’s costs plus a markup. But in recent years, venture capital has upended that common sense with a push for scale and worry about profits later. We don’t know whether Starlink, like Uber and others before, are deliberately discounting prices to capture the market.

Starlink would not be the first Silicon Valley initiative to tout its financial sustainability, only to fall like a house of cards when the real accounting begins. We need look no further than Alphabet’s Project Loon, an ambitious initiative to create a global fleet of broadband balloons, to see a cautionary tale of tech solutionism. In 2020, Loon and its like were projected to generate $4B in revenue within 10 years. In 2021, Project Loon was abruptly shut down by Alphabet.

I feel pretty confident in saying that Starlink’s current pricing has little to do with their costs but it’s worth trying some back-of-the-napkin maths to probe this further. Here are my assumptions about Starlink, some documented, some guesses:

  • A launch to low earth orbit on a Falcon9 rocket costs SpaceX about US$30M. Starlink have said as much and they charge others US$67M for the same, so that seems plausible.

  • A SpaceX Falcon 9 rocket could lift about 55 of the 1st generation satellites (260kg each) into orbit. However, Starlink has moved on to the heavier (800kg) V2 mini satellites, which means that they can only launch about 22 satellites each time. The V2 mini satellites are said to have about 4 times the capacity of the V1 but as the size of the constellation remains the same, this doesn’t affect the calculations.

  • Starlink is planning to have a constellation of 12,000 satellites. The target constellation started out at 4425 satellites and, on October 25th, Starlink applied to launch 30K additional satellites. However, the current constellation plans are for 12,000 satellites so that is the number I’ll go with.

  • The average lifespan of a Starlink satellite is about 5 years. Being so close to the earth means that Starlink can offer low-latency connectivity but also means that Starlink satellites are more affected by drag caused by the earth’s atmosphere. Low earth orbit satellites need to be periodically boosted back into their orbits, thus their lifespan is constrained by the amount of fuel they can carry to maintain their orbit.

  • Starlink’s failure rate has been estimated at 2.5 to 3%. However, tracking Starlink satellites has suggested that of the 5331 Starlink satellites that have been launched, 363 are not operational, leaving a total of 4968 active satellites. This makes the failure rate closer to 7%. Who knows? In order to keep my calculations as simple as possible, I am going to leave this factor out for now but it is worth keeping in mind.

So what does this tell us? Assuming all satellites will be V2 mini satellites as the 1st generation expires, each satellite costs ~$1.36M ($30M ÷ 22) to launch. I don’t know how much a Starlink satellite costs to manufacture but I am going to choose the lowest estimate I have seen, which is about $250K. Let’s call it ~$1.6M ($1.36M + $250K) per satellite per launch.

That means that it will cost about ~$19B to build a constellation of 12,000 satellites. And here’s the kicker. The moment those satellites go up, they are on a 5 year clock. In order to maintain this constellation Starlink must launch 2,400 satellites per year just to keep 12,000 satellites in orbit. That is nearly $3.9B per year just to maintain the constellation. And this is not taking into account all of the other costs that Starlink incurs, ranging from 150 ground stations (with many more planned), to R&D and manufacturing costs, to national licensing costs, to the small army of lobbyists that span the globe seeking licensing approval for Starlink services.

Starlink has claimed that launch costs will drop dramatically when the much larger Starship rocket comes into operation. However it is worth noting that Starship is an extraordinarily ambitious undertaking that has yet to complete a successful launch. Furthermore, it is the subject of a lawsuit initiated by a coalition of environmental groups as a result of its first launch attempt. All that to say that Starship is currently too large an unknown to be able to factor into Starlink sustainability.

As things currently stand, Starlink made $1.4B in revenue in 2022. They had projected to have 20 million subscribers by 2022 but as of Sept 2023 they had 2 million. They have quite a long way to go before they are anywhere near breaking even. 

Some may argue that a combination of military contracts and government subsidies will keep Starlink alive and this may be true. However, given the capriciousness of Starlink’s owner, Elon Musk, governments may desire military satellite constellations to be under their complete control. And indeed, that is what is happening.

Furthermore, the sweet spot of connectivity for Starlink is surprisingly small. Their target markets are comparatively wealthy rural dwellers who live in remote or sparsely populated areas in which big telcos don’t see a sufficient return for investing in terrestrial infrastructure. Starlink can’t compete in urban areas because there are typically already competitive services available and satellite constellations can’t cope with too many subscribers per satellite, not to mention the need of satellite terminals to have a wide view of the sky. Their key market is always under pressure as telcos do eventually find ways to expand their infrastructure deeper into rural areas.

This year, Starlink made much of receiving permission to operate in several African countries. However, their business case is even weaker in this region. While there is a desperate need for rural connectivity, the number of people living in rural areas with the ability to pay Starlink’s hardware and monthly cost is quite low. For example, as of 2022, an estimated 40 million Nigerians have an income level below the national poverty line of 137,430 naira per person per year, equating to less than US$2 per day.

In theory, Starlink could provide a very useful service to rural Internet Service Providers (ISPs) offering a high speed internet backbone for rural telephony or internet services. But, to do that profitably, Starlink would have to charge a lot more money, which again makes them a less attractive option. And again the number of ISPs that might be Starlink clients is not very large.

In sum, it is impossible for the general public to know what Starlink costs to operate. In November 2023, Elon Musk claimed that Starlink had achieved breakeven cash flow. There is no way to interrogate that claim. Starlink may look profitable, but only as long as Elon Musk is prepared to pour money into it. And, the basic numbers above paint a different picture. I make no claim to authority on this and welcome anyone to improve or correct my somewhat simplistic assumptions.


I am grateful to Carlos Rey-Moreno, Peter Bloom, Adriana Labardini, and Katherine Barrett for invaluable feedback on the original draft. All errors and underlying assumptions remain my own. This article presents my personal perspective and does not represent the views of Mozilla or APC.

This piece was originally published on the author's blog Many Possibilities. It has been split into two parts for republication, and you can read the second part here.

Image: Starlink Mission by Official SpaceX Photos via Flickr (CC BY-NC 2.0 DEED)

Steve Song is a consultant on access regulation and policy to APC's Local Networks initiative and a policy advisor with the Mozilla corporation. His blog, manypossibilities dot net, is a popular destination for anyone working on African telecommunications and internet issues.
APC-wide activities: 
« Go back