Three high satellites versus more than 11,000 low ones
A geostationary satellite (GEO) orbits at 35,786 km, above the equator, and takes 23 hours 56 minutes to go around once. That matches Earth's spin, so it seems to hang in one spot in the sky [1]. From that height, three evenly spaced satellites can cover nearly the whole globe [1].
Low Earth orbit (LEO) is the region below about 2,000 km. NASA's Earthdata team gives it as roughly 160 to 2,000 km [1] [2]. Starlink's first batch, approved by the US Federal Communications Commission (FCC) in April 2021, had 4,408 satellites. They were arranged in layers, called shells, at heights of 540 to 570 km, with each shell tilted at a different angle to the equator [3].
How many are up there now? Astronomer Jonathan McDowell keeps a public count. On October 15, 2026, he listed 13,015 Starlink satellites launched in total, 11,154 still in orbit, and 11,140 counted as working [4]. The numbers differ because some satellites have fallen back to Earth, and a few have failed. This article rounds to "more than 11,000".
We cannot say "three versus 11,000, so Starlink is 3,700 times bigger." The two systems are built to do different jobs. What we can ask is why anyone would give up the three-satellite design.
Why a satellite 36,000 km away makes the signal wait
Radio waves travel at the speed of light, about 299,792 km per second. That is fast, but not instant. A signal that goes up to a geostationary satellite and back down travels about 2 × 35,786 km. Divide by the speed of light and you get about 0.24 seconds. This is my own calculation, and it assumes the satellite is directly overhead.
Loading a web page needs a request and a reply, and each one makes the trip through the satellite. That doubles the time, to about 0.48 seconds at the very least. Real systems add processing time, ground-station links, and busy-network delays, so the real wait is longer. This lower limit comes from distance alone, and no better equipment can shrink it.
Moving closer cuts the delay by about 65 times
Now use the same formula at 550 km, the typical height of Starlink's first shells [3]. Going up and down takes 2 × 550 ÷ 299,792, which is about 0.0037 seconds, or 3.7 milliseconds. A request and a reply make about 7 milliseconds. Since 35,786 ÷ 550 is about 65, the lower limit shrinks by the same factor.
Measured delays are bigger than these limits, because they include more than distance. The company Ookla measures internet speeds from users' tests, and the IEEE Communications Society's blog reported its median delays for the first quarter of 2025: about 45 milliseconds for Starlink, 683 for HughesNet, and 684 for Viasat. The last two use geostationary satellites [5]. Those are real-world medians, not physics limits, so they cannot be compared directly with the calculated floor. They do show the same direction: the nearer system is about 15 times quicker.
Why a satellite close to Earth passes overhead so quickly
There is a cost to flying low. The lower an orbit, the faster the satellite must move to stay up. ESA says LEO satellites go around in about 90 minutes at roughly 7.8 km per second [1]. Using Kepler's third law with Earth's gravity constant and radius, I worked out about 95.6 minutes and 7.6 km per second for a 550 km orbit, a result that agrees well with ESA's numbers. A geostationary satellite moves at only about 3 km per second [1].
A low satellite also sees a smaller patch of ground. Think of looking at a city from a drone: the lower it hangs, the less of the city it sees. So a single Starlink satellite rises over the horizon, crosses the sky, and sinks below the other side during a small part of its lap. A dish on the ground cannot stay locked onto it.
That is why a low-orbit system is useless with only one or two satellites. Near-continuous service needs a large number, arranged so that another satellite is always rising when one is setting.
How a constellation of satellites works as a relay team
A group of satellites run together as one system is called a satellite constellation. As one satellite drops toward the horizon, the connection is passed to the next one that is rising. Engineers call this a handover. It is like a relay race: no runner needs to be fast for the whole race, as long as the baton keeps moving.
Layers help too. Starlink's first 4,408 satellites were spread over four shells with different heights and tilts: 1,584 at 550 km (53 degrees), 1,584 at 540 km (53.2 degrees), 720 at 570 km (70 degrees), and 520 at 560 km (97.6 degrees) [3]. Different tilts reach different latitudes, and tilts near 90 degrees pass over the polar regions. That last point is a simple way to picture it, not something the FCC report spelled out.
Low flight also affects safety. The FCC's 2021 decision brought more than 2,800 satellites down from planned orbits as high as 1,300 km to 540 to 570 km. SpaceX argued that this reduces the risk of dead satellites staying in orbit as space junk [3]. That is the company's argument, not something we have tested.
Check the delay yourself with a calculator
You need only two inputs: the height of the orbit and the speed of light. Type 35,786 × 2 ÷ 299,792 and you get about 0.24. Then type 550 × 2 ÷ 299,792, and you get about 0.0037. Why are the answers so different, and what is the price of getting the small one?
If you want to dig deeper, ESA's "Types of orbits" page explains GEO, LEO, and other orbits. McDowell's page lists every Starlink launch, so you can watch the count change. Some questions remain open here: how many low satellites are needed to cover one region, exactly how handovers are controlled, and how well retired satellites really burn up on the way down. Those are good questions to chase.