The megaconstellation decade

For sixty years, a satellite was a cathedral: designed once, built by hand, launched alone, and expected to work for decades because there would never be a replacement.

That era built wonders, from the geostationary ring to the GPS constellation, but it kept orbit sparse. The whole planet was launching a few hundred objects in a good year, and the population in orbit crept upward at roughly the same pace for half a century. The fleets that did exist were small and precious: GPS has never needed more than a few dozen satellites, because from 20,655 kilometres up each one sees nearly half the planet at once.

The first hint of a different model came in 1997, when Iridium began launching the first production-line constellation: 175 satellites ever built to one design, enough to put a phone signal over every point on Earth. It was a commercial heartbreak at the time and a technical triumph in the long run, and it proved the recipe every modern fleet now follows.

Three curves crossed in the late 2010s. Rockets became reusable, which cut the price of a kilogram to orbit several times over. Satellites became flat: Starlink's design stacks like sheets of plywood, so a single launch delivers dozens at a time instead of one. And consumer electronics made the payload cheap, with phased-array antennas and mass-produced components replacing bespoke space hardware.

In 2019 both OneWeb and Starlink launched their first operational batches, and the catalog bent upward within months. The numbers stopped resembling the old space age almost immediately: in under a decade Starlink alone has flown 12,672 satellites, more than everything the whole world launched in the space age's first fifty years combined.

The picture below is the whole story in one frame. For six decades the population in orbit is a slowly thickening grey band of one-of-a-kind spacecraft and spent rocket bodies. Then the coloured fleets arrive, and within a few years named constellations make up 57% of everything in orbit.

Fleets this size changed the rhythm of spaceflight itself. Launch was the bottleneck of the old space age; it is now a conveyor. The same booster flies, lands and flies again within weeks, and a constellation mission is less an event than a delivery run.

The catalog puts numbers on it: in the half century before the fleets, the record year was 1998, when 652 tracked objects went up. In 2025 it was 4,619.

The named fleets the atlas tracks split into two families. The broadband constellations fly low and count in the hundreds or thousands; the navigation constellations fly at twenty thousand kilometres and count in the dozens, because altitude buys coverage. Every row links to the full fleet list, and the reentered column is worth reading closely: for the low fleets it measures retirement working as designed.

A geostationary satellite serves a continent from one fixed point, so why build thousands of small satellites instead of a few big ones? The answer is the speed of light. A signal to the geostationary ring travels 36,000 kilometres each way, and the round trip adds a quarter of a second before the internet even starts responding. From Starlink's shells around 469 kilometres the same round trip is measured in milliseconds, which is what makes video calls and gaming over satellite feel like cable.

Low altitude sets everything else about the design. Each satellite sees only a small patch of ground and crosses the sky in minutes, so coverage must come from numbers: dozens of orbital planes, dozens of satellites per plane, handing users off continuously. The inclinations in the table are the fingerprint of that geometry. Starlink's main shells at 53 degrees blanket the latitudes where most people live, later shells at 43 and 97 degrees fill in the tropics and the poles, and OneWeb went straight to a polar 88 degrees at 1,197 kilometres so a smaller fleet could reach the whole globe at once.

The obvious worry about ten thousand new satellites is debris, and the industry's answer is built into the orbits themselves. The big broadband shells sit low enough that atmospheric drag is a cleaning service: a satellite that dies silently is dragged down and burns up within a few years, no intervention required. That is a deliberate design choice, and it is the single biggest difference between the new fleets and the debris that older, higher orbits still carry.

Healthy satellites do not wait for the atmosphere. At end of life a Starlink satellite uses its ion thruster to steer itself down, and the catalog shows the system working at scale: 1,695 Starlink satellites have already reentered, retired deliberately as newer generations replace them. The fleets also carry automated collision avoidance, screening every close approach and moving out of the way, and regulators have kept pace, with the old twenty five year disposal guideline now tightened to five for low orbits.

None of this makes traffic management a solved problem; tens of thousands of satellites sharing the low shells will keep demanding better coordination, tracking and rules. But the trajectory is encouraging: the busiest decade in spaceflight history has also been the one where deorbiting on schedule became normal engineering practice.

Megaconstellations are bright enough to matter to astronomy, and the first Starlink trains in 2019 startled everyone, streaking through long-exposure images of the deep sky. What followed is a better story than the headlines suggested: operators and astronomers sat down together, and the satellites changed.

Newer generations fly with darkened surfaces, sun visors and off-pointed solar panels during twilight, cutting their brightness several times over, and orbital positions are published so observatories can schedule around passes. The satellites remain visible to careful observers, especially in the days after launch when a fresh train climbs to altitude, and that sight has become its own small phenomenon: for many people a Starlink train is the first spacecraft they ever notice with their own eyes.