How Starlink Keeps A Laser On A Moving Satellite

Watch SpaceX's footage of its first V2 Mini deployment, from February 2023. Flat satellites peel away from the rocket one after another. In Starlink's low orbits, each of them ends up moving at about 17,000 mph, roughly 27,000 km/h.

Now hit one of them with a laser. Fire it from another satellite moving just as fast, and keep the beam on target long enough to push internet traffic through it.

That sounds absurd. I think the 17,000 mph figure is the wrong number to worry about, and once you see why, the whole laser network makes sense.

The number that misleads you

For aiming, what counts is how the other satellite moves across your view. Its speed around Earth tells you very little about that.

Picture yourself as a passenger in a car. Another car drives beside you at the same speed. Look out the window and the other passenger barely moves. The road streams past while the person next to you stays put. Same trip, two completely different views.

Now hold a flashlight toward a receiver in that other car. Your car is the transmitting satellite. The other car carries the receiving satellite. The beam is the link. Reading your own speedometer won't tell you where to point. You need the other car's motion relative to yours.

In orbit, take an idealized pair of satellites on the same circular path, one ahead of the other, both going around at the same pace. They stay the same distance apart. With Earth staying below you, the neighbor can sit at a fixed angle in your view. A spacecraft doing 17,000 mph can look, from where you sit, like a target that holds still.

Change the paths and you change the job. Satellites on crossing paths sweep across each other's view. The direction between them keeps shifting, and their window to connect can be shorter. From the passenger seat, someone riding beside you is easier to follow than someone cutting across your path.

Why the beam stays tight

A fixed direction still has to be the right direction, and a narrow beam makes any mistake expensive.

Go back to the flashlight. Spread its light over a huge area and only a sliver lands on what you're trying to light up. Concentrate it and more of it reaches the target. That matters when the receiver is far away and needs enough light to read the data.

The same choice creates the problem. Twist a tight flashlight a little and the bright spot slides off the receiver. Put the receiver farther away and that same small twist moves the spot a greater distance. The error grows as the light travels.

And a lot is coming in. In January 2024, SpaceX described its laser links as 100 gigabit links that could peak at 200, and Starlink now rates each laser at up to 200 gigabits per second. That's as many as 200 billion ones and zeros crossing the gap every second, and the receiver has to keep catching usable light for the entire transfer.

Predict, acquire, track

Before you can keep a beam centered on a receiver, you have to find it. Knowing roughly where it sits doesn't mean your narrow beam has reached it. Finding a partner and staying connected are separate jobs for the laser terminal, the hardware that sends and receives the light.

The job starts with a prediction. Where should the other spacecraft be, and which way should the terminal point to reach it? That gives you somewhere to look. But the direction you intend and the direction the hardware actually points differ slightly. The way the spacecraft is facing affects the aim. So does where the terminal is mounted on it. A small error in either shifts where the light goes.

Next comes acquisition: getting both ends into a position where light from the other terminal can actually be detected. Then tracking: once the light is there, keep following it as the geometry changes.

How does a machine know its aim is drifting? It measures what arrives. A light sensor reports the direction of the incoming light, and the control system uses that reading to correct the pointing. That's feedback, and it keeps running while data flows. The prediction tells the terminal where to look before it has seen any light. The measurement tells it what is really arriving, so it can fix small errors as they happen.

NASA's terminal on the space station

NASA ran this sequence on the International Space Station, and it gives you real hardware to picture. The equipment went up in November 2023, and its experiments ended in June 2024. The station sent light to another satellite, which relayed the information onward: a moving spacecraft holding an optical link to a partner in space.

The terminal started from the station's position, the direction the station was facing, and a prediction of where the relay satellite would be. Combine those and you get a starting direction before the other end has seen any of your light. That starting direction still carried some error, so the terminal sent out a wider beam to help the relay find it. That wider beam is called a beacon. Giving the light more room makes first contact easier, and once the far end catches it, there's something real to measure.

Then the moving parts. The terminal sat on a mount that pivoted in different directions, and it carried fast-moving mirrors. The mount swung the optical equipment around. The mirrors handled the finer adjustments to where the beam went. A light sensor found the incoming light and kept tracking it, and the control system turned those readings into pointing corrections.

Aim where it's going

Even a terminal with zero shake has one more problem. Light is fast, but crossing the gap still takes time, and the receiving satellite keeps moving while the beam is in flight. The beam has to meet the receiver where it will be when the light arrives. Aim at the spot it just left and you miss.

The correction is called point ahead: aiming ahead of a moving target. How much lead you need depends on the sideways motion across the beam's path during the trip. More sideways motion means more lead. Once again, orbital speed by itself doesn't tell you how far ahead to aim.

There's a second wrinkle. Light coming in from the other satellite also took time to travel, so when it reaches your sensor it shows where that satellite was a moment earlier. The terminal is juggling two directions at once: receive light that left a past position, and send light toward a future one. In this geometry, following the incoming beam and choosing where to fire the outgoing one point in different directions.

The answer is math. Use the predicted orbital motion and the time light needs to cross the gap, and you can work out where the receiver will be when your beam lands. That adjustment has to keep up as the geometry changes, and a pair flying together calls for a different calculation than a pair crossing each other's view.

Sixteen burns, one link

All of this has to work on an actual spacecraft. Beyond its path around Earth, a vehicle shifts in small ways. Think of the passenger with the flashlight again: the other car stays beside you, but your hands move. On a spacecraft, small changes in orientation do the same thing to the beam.

In September 2024, the Polaris Dawn crew tested a laser connection between their Dragon spacecraft and Starlink, and SpaceX shared pictures of the crew talking with its teams over that link. On September 14, SpaceX posted that the laser in Dragon's trunk and its partner Starlink satellite held their link through 16 Draco thruster firings.

Weeks per link, a quarter million handshakes a day

A demo proves the aiming. A network has to carry traffic hour after hour while everything keeps moving, so endurance is the real test.

In January 2024, SpaceX engineer Travis Brashears presented results at SPIE Photonics West, a laser conference in San Francisco. By then SpaceX was describing laser links held for weeks. That fits the good-neighbor idea from earlier: when two satellites' paths keep the geometry manageable, their terminals can use the same connection for a long stretch.

The fleet also has to keep finding new partners. The same presentation put that at about 266,141 laser acquisitions per day, more than a quarter million across the system. Those 2024 results also put link uptime above 99%. Long-lived pairs plus constant new acquisitions: that's how the network keeps usable connections available as it moves.

The planet gets in the way

Put two satellites on opposite sides of the planet and a laser between them runs straight into Earth. No amount of aiming fixes that. Every direct link needs a clear line of sight.

Distance creates a sneakier version. Both satellites can be above the atmosphere while the straight line between them dips into it. Stretch that line across the curve of Earth and its middle passes closer to the surface.

That happened on a real Starlink link. At the January 2024 presentation, Brashears described connecting satellites more than 5,400 km apart, about 3,355 miles. The path dipped into the atmosphere, and the connection eventually broke.

For long distances, the fix is more satellites along the way. The first sends data to a partner it can reach. That partner passes it farther. Each link covers its own stretch, and traffic works around the planet one segment at a time.

Starlink's 2025 progress report describes three laser links on each V2 Mini across the fleet. Three links per satellite build a mesh, a network with multiple paths between places. The payoff for you is choice of route. The satellite above you doesn't need a direct laser line to every place your message is headed. It needs a way into the wider network and a path that eventually reaches the ground.

The long way home from a ship

Take a hypothetical. You're on a ship and send a message home. Your dish sends it up to a Starlink satellite by radio. Now suppose that satellite has no suitable gateway in view, meaning no ground station linking the satellite network to the internet.

So it hands the data to another satellite over a laser link. From there the data hops through the mesh until it lands on a satellite with a gateway below it. That last leg drops back to Earth by radio, and the internet carries the message the rest of the way.

Up by radio, across by laser, down by radio. That's the entire route.

Antarctica shows it working. In September 2022, SpaceX announced Starlink service there, and its post said: "In such a remote location like Antarctica, this capability is enabled by Starlink's space laser network." A user reaches the satellite overhead, and the satellites carry the traffic toward a ground connection the first one can't reach directly.

Routes change too. In January 2024, Brashears described switching routes within milliseconds. That's the network choosing another path through laser links already in place. Finding a new optical partner and picking a path through existing partners are separate jobs. The chain from your ship to the ground can change while your message keeps heading for the same destination.

Glass versus vacuum

Once the route runs through space, the speed of light starts to matter. Fiber already moves data as light, but light travels faster through a vacuum than through the glass in ordinary fiber cable. Take two paths of equal length, one through that glass and one through empty space, and the light finishes the empty one first.

For a nearby destination with a direct fiber connection, sending data into space is a detour. Even inside Starlink, dropping straight to a suitable ground station can be quicker than taking extra laser hops. Over long distances the balance shifts. More of the trip benefits from faster light in space, and a satellite route can skip a long detour through cables on the ground.

There's a real example. In a July 2024 analysis, researchers looked at Starlink service on Réunion Island. Its traffic reached an internet connection point in Germany, and the time to send data and get a reply was shorter than with the ground-based service they compared it against. The laser links let Starlink route traffic directly between satellites, while the cable routes available from the island took long detours.

What all that aiming buys

I think this is why anyone would take on something this hard. A single satellite only reaches users and ground stations it can see. Laser links stretch your connection past that, so a ship far offshore or a station in Antarctica can reach a gateway nowhere nearby. The tiny aiming problem connects directly to whether somebody's internet works.

For the person on that ship, none of it shows. They send something and it arrives. That's a lot of engineering spent so a message home can feel ordinary.