Free topic

Traffic jam in orbit

by Joe Miller

Low Earth orbit went from about 2,000 working satellites in 2019 to about 16,900 today. SpaceX wants up to a million more for AI data centers, and China is building two networks of more than 13,000 satellites each. Here's how crowded it's getting, what could go wrong, and what to expect.

Watch: the narrated video (6 minutes)

How crowded it is Who is launching Who flies where Constant dodging The chain reaction The rules What to expect Solutions Sources

Plain-English explainer by Joe Miller, a nuclear engineer. Not investment advice. Figures are the latest public numbers we could find and change fast; sources are at the bottom of each page.

How crowded it is

Drawn to scale: the data center plan is about 60 times everything in orbit today.

About 16,900 working satellites are in orbit, and about two-thirds of them are SpaceX Starlinks.
Counting dead satellites, old rocket stages and broken pieces, the European Space Agency tracks about 47,000 objects, plus an estimated 1.5 million pieces between half an inch and four inches across, too small to track.
Everything up there moves about 17,000 miles per hour. At that speed, a piece the size of a marble hits with about the energy of a car crashing at 30 miles per hour.
The world made 329 orbital launches in 2025, a record, and 268 through mid-November 2026. SpaceX did nearly half of them; China about 30%.

Who is launching

Network	Who	In orbit now	Planned
Starlink	SpaceX (U.S.)	about 11,200	15,000 approved; more requested
Starmind data centers	SpaceX (U.S.)	none yet	up to 1,000,000 (applied Jan. 2026)
Qianfan ("Thousand Sails")	Shanghai Spacecom (China)	about 250 to 270	more than 15,000
Guowang ("national network")	China SatNet (China)	about 220 to 240	about 13,000
Amazon Leo (formerly Kuiper)	Amazon (U.S.)	about 390	3,236
OneWeb	Eutelsat (Europe)	about 650	replacement fleet planned
Rassvet	Bureau 1440 (Russia)	about 30	about 900
IRIS²	European Union	none yet	348

Counts as of October to November 2026 from KeepTrack, Orbital Radar and azmth.space; they change weekly.

China's two networks are its answer to Starlink, and they are launching steadily: together about 500 satellites so far, with full networks aimed at around 2030 and beyond. Qianfan had satellite failures in 2025 that paused launches for several months before they resumed in April 2026.

Who flies where

Height matters: below about 600 km, thin air drags dead satellites down within years. Above about 1,000 km, they can stay for decades or centuries.

This is the part most people miss. Starlink flies low, around 300 to 350 miles up, where a dead satellite falls back and burns up within a few years. Much of China's Guowang and Qianfan fly higher, around 500 to 700 miles up, and SpaceX's data center plan reaches up to about 1,240 miles. Junk left that high doesn't clean itself up in our lifetimes.

Constant dodging

Every Starlink satellite now dodges something about once a week.

Starlink made more than 355,000 collision-avoidance moves in the year to May 2026, more than three times the 2024 total. Its satellites steer automatically when the chance of a hit passes about 3 in 10 million.
Close calls with China: after a Starlink satellite reportedly passed within about 200 meters of a Chinese satellite, SpaceX began lowering about 4,400 Starlinks from about 340 to about 300 miles up during 2026, which also makes failed ones fall faster.
The math gets worse fast. Doubling the satellites in a band roughly quadruples the number of pairs that could collide. One researcher projects more than a million Starlink dodges a year by 2030, and each dodge still leaves a tiny risk that adds up.
No common traffic system. SpaceX, China and others each run their own tracking. Companies share some data, but there's no single air-traffic-control system for orbit, and the U.S. and China have limited channels for coordinating.

The nightmare: a chain reaction

In 1978, NASA scientist Donald Kessler warned that orbit could reach a tipping point: one collision makes thousands of fragments, those fragments hit other satellites, and the debris keeps multiplying on its own. It's now called the Kessler syndrome. It wouldn't happen overnight like in the movies; it would build over years and could make some altitudes too dangerous to use. The Outer Space Institute estimates that if every satellite stopped dodging today, a major collision would be expected within about four days. That's how much we already depend on constant steering.

Real breakups have already happened: a 2009 collision between a U.S. Iridium satellite and a dead Russian satellite, and anti-satellite missile tests by China (2007) and Russia (2021) that created more than 3,000 and about 1,500 trackable fragments.

The rules today

The FCC's five-year rule (2022): U.S.-licensed satellites in low orbit must be brought down within five years after their mission ends. SpaceX says it brought down 260 old Starlinks in the six months to May 2026 and that more than 99% of its retirements work as planned.
International guidelines are voluntary. The United Nations has guidelines, but no worldwide rule with teeth covers every country's satellites.
SpaceX asked for a break on its million-satellite plan: FCC rules normally require half of a network to be in orbit within six years of approval, and SpaceX asked to waive that.

What to expect

Traffic will keep growing no matter what. Starlink, Amazon and China's two networks alone add up to more than 45,000 planned satellites before SpaceX's data center plan.

The million-satellite plan, if approved at all, will likely come in stages, a smaller first batch with conditions on altitude, brightness and disposal, with more only if the first ones behave. The FCC hasn't ruled.

Solutions: what would actually fix it

Solution	What it means	Where it stands (Nov. 2026)
1. Air traffic control for space	One shared system that tracks every satellite and piece of junk and warns operators of close calls, used by every country.	The U.S. runs a pilot, TraCSS, with 62 operators and more than 11,000 satellites, but its 2027 funding is in doubt. China shares little data.
2. Keep big fleets low	Require megaconstellations to fly below about 600 km, where dead satellites fall out within a few years.	Starlink is moving lower. Much of China's networks and SpaceX's data center plan would fly much higher.
3. Faster cleanup rules	Bring dead satellites down within 1 year instead of 5, prove disposal works, and empty leftover fuel from rocket stages.	The U.S. has a 5-year rule. Worldwide rules are voluntary.
4. Approve in stages	License big networks in batches, with more allowed only after the first satellites prove reliable. Make operators disclose expected dodges and fuel.	The FCC can do this today for the million-satellite plan.
5. Make polluters pay	A small yearly fee per satellite, or a bond refunded when a satellite is safely removed, so crowding has a price.	Proposed by economists; not adopted yet.
6. Remove the worst junk	Robot spacecraft grab and drag down dead rockets and satellites, starting with the biggest and most dangerous.	Japan's Astroscale got within 15 meters of an old rocket stage; a removal mission is aimed at 2027-28. No junk has been removed from low orbit yet.
7. No more missile tests	A worldwide ban on anti-satellite tests that blow satellites into thousands of pieces.	The U.S. and dozens of countries pledged to stop; China and Russia have not.
8. A U.S.-China hotline for space	Direct operator-to-operator contact so a close call is a phone call, not a guess.	Limited today. Close calls are mostly handled by each side dodging on its own.

Sources at the bottom of the page.

How we get there

Now, in the U.S.: the FCC attaches conditions to big licenses, including the million-satellite plan: low altitudes, approval in stages, proof of disposal, and brightness limits.
Fund the traffic system: Congress keeps TraCSS running, and satellite operators may share the cost through user fees.
Bring in everyone: the U.S., Europe, Japan, China and Russia agree on shared tracking data and rules of the road through the United Nations space committee and the radio-spectrum body, the ITU.
Price the risk: insurers and governments charge more for satellites that are risky, high or hard to remove.
Clean up: pay for removal missions to bring down the 50 most dangerous dead objects first.

What you can do: FCC filings are public, and anyone can comment. When the FCC asks for comments on big satellite plans, short, specific comments about altitude, cleanup and approval in stages do get read.

My take: in a nuclear plant, we never rely on everything working perfectly; we design for the failure. Orbit today relies on hundreds of thousands of perfect dodges a year. That works until a satellite dies at the wrong altitude or two operators don't talk. The million-satellite plan makes that bet much bigger. See Elon Musk, Colossus and data centers in space for why SpaceX wants it.

Members: $15 a month

Half the topics are free. Members get the other half, plus the monthly Data Center Critical newsletter with the month's AI, chip and power news. Cancel anytime.

Join for $15/month

Who's writing this? I'm Joe Miller, a nuclear engineer. I spent my career around power plants, so when the AI boom turned into a race for electricity, I started digging. This site explains what I've found in plain English. My story is at criticalmassbook.com.

Sources

Checked November 14, 2026.