Historical Overview

LEO constellations did not appear yesterday. This is a story of gradual waves, in each of which ambitions, money, and technologies clashed. When we see three names today – Starlink, OneWeb, and Kuiper – it is important to remember that they did not start at the same time and with the same resources. These are different generations of the same race.

OneWeb started earlier than everyone else. It became the first “big” project of the second wave of satellite internet. Its first satellites appeared in orbit when Starlink was still in the presentation stage. But the technology laid down in Gen1 quickly turned out to be outdated: the satellites are light, without optical inter-satellite links, with fixed beams, and strictly dependent on a dense network of gateways. This choice was logical in the mid-2010s, but against the backdrop of subsequent progress, it looks like a limitation. Today, OneWeb is more trying to “seamlessly” transition to the next generation than to compete on an equal footing with newer architectures.

Starlink appeared a little later but immediately in a different league. First, SpaceX had its own rockets and could launch a hundred satellites in one go, removing the main barrier – the cost of access to orbit. Second, from the very beginning, the network was built on the ideas of scalability and upgradability: from Gen1 with simpler terminals to Gen2 with laser links and a wider spectrum. This created the effect of a “mass breakthrough” – Starlink became synonymous with satellite internet. But rapid growth also brought its own problem: today, SpaceX simultaneously operates several generations of satellites with different capabilities. This is both flexibility and a major headache.

Kuiper is the youngest of the trio. Amazon started later but used the lessons of both predecessors. Three altitudes were immediately announced, optical links on board, its own Prometheus ASIC with terabit processing, and integration with the AWS cloud. Where OneWeb bet on partners, and Starlink – on the speed of mass launches, Kuiper from the beginning is building a vertically integrated ecosystem: from chips and terminals to data centers and client services. In some respects, it is even more modern than Starlink – precisely because of the “late start.”

Thus, the three systems are not just competitors on the same playing field. They are three stages of one story. And to understand their current strengths and weaknesses, you have to take this temporal context into account: who started earlier, what compromises were inevitable then, and how they now affect the architecture and capabilities of the networks.

Why you can’t just count satellites

“We will have ten thousand satellites!” – that’s how headlines sound, which are easy to be misled by. To a non-specialist, everything looks simple: more satellites = better network. But in satellite internet, arithmetic doesn’t work directly. The key question is not how many metal boxes are flying overhead, but what each of them can do and how they are stitched together.

Imagine two scenarios. In one – you have hundreds of light satellites, each with fixed beams, no laser links, and a limited ability to transmit traffic further than to the nearest ground station. In the second – a smaller number of heavier satellites, but with powerful phased array antennas, flexible beam-hopping, and inter-satellite channels at hundreds of gigabits. The first network on paper may look more massive, but during peak hours in a metropolis or over an air route, the second will handle the load better, even with fewer “bodies” in the sky.

There are at least four reasons why simple counting is misleading:

• Beam density is more important than the number of satellites. A single satellite with a hundred narrow beams can serve thousands of users in a city better than a few “old” ones with a dozen wide ones.
• Ground infrastructure creates bottlenecks. If there are not enough gateways or powerful inter-satellite links, additional satellites will not help – traffic will hit a “bottleneck.”
• Spectrum is a limited resource. The FCC and ITU regulate power and frequencies. Additional satellites will not expand the bandwidth if it is already exhausted.
• Distribution policies and network generations. A network with different versions of satellites may have compatibility issues. Sometimes this even limits capabilities, instead of increasing them.

So, the “number of satellites” is more an indicator of the scale of production and launches than of service quality. To really understand whose network is more powerful, you need to look at the orbital architecture, antenna flexibility, spectrum, and how the system copes precisely where demand is highest. And that’s exactly what we’ll analyze next.

Orbital Architecture

When they say “LEO,” a single “low orbit” is imagined. In fact, it is a whole set of altitudes and inclinations that determine how satellites will “hug” the Earth. For the user, these are imperceptible details, but they are the ones that determine whether they will have a stable connection over Paris, the Atlantic, or in northern Finland.

OneWeb chose an altitude of about 1200 km and almost polar inclinations. This means global coverage, including Arctic regions. But altitude comes at a price: higher signal latency and fewer opportunities for spectrum reuse. And most importantly – in the first-generation satellites (Gen1), optical links are absent, so the entire network depends on ground gateways. On paper, it looks beautiful: a full “cap” over the planet. In practice – a huge dependence on where and how many ground stations are actually built.

Starlink went a different way: lower altitudes – about 530–570 km – and several shells with different inclinations. This gives lower latency and better beam density in “hot zones.” But at the same time, it creates a new challenge: SpaceX simultaneously operates several generations of satellites with different sets of capabilities. Some are without ISL, some are with them. The company is already planning even heavier satellites of the next generation (V3) and has even applied for more than 14 thousand satellites for direct communication with smartphones (Direct-to-Cell). This means that the fleet is turning into a complex zoo, where management and updating require enormous resources. SpaceX’s bet from the very beginning is not just presence, but domination. But the scale and speed forced them to make compromises, and where there is a compromise, there is always an opportunity for competitors.

Kuiper entered the game later and immediately bet on a multi-layered approach: three shells – 590, 610, and 630 km. This architecture allows for flexible balancing of coverage and capacity, reinforcing “hot latitudes” and at the same time avoiding excessive duplication of trajectories. And unlike OneWeb, Kuiper from the start equips satellites with optical links, which means traffic can travel “in the sky,” bypassing scarce gateways.

Altitude, inclination, the number of planes – these are not abstract numbers. They determine the “geography of service availability”: whether you can rely on the network on a polar flight, how low the latency will be, and whether there will be enough resources in a metropolis during peak hours. And this is where it becomes clear how different the approaches of the three players are – from the outdated Gen1 in OneWeb, to the complex multi-generational Starlink fleet, to the modern multi-layered Kuiper architecture.