Nine of 16 planned satellite fleets are too big for orbit

A model of 16 real satellite fleets found nine of them exceed the size at which their own wreckage keeps growing. Six of those nine cross the worst line, the runaway one.

What the paper found

Hugh G. Lewis built a stability model that gives each satellite fleet a critical size. He ran 16 real planned or deployed fleets through it. Nine of the 16 sit above their critical size. Six of those nine would produce wreckage that grows without limit, even in an empty sky and with better than standard clean-up rules.

By Mark Alex · September 9, 2026

Grade 5 reading level5 min readPreprint · not yet peer reviewed

What happened

Picture a car park with one exit. A few cars leave with no trouble. Fill it with ten thousand cars, and one small dent starts a jam that never clears. The space just above your head is close to that point.

Satellites fly there in big rings. A ring is called a fleet. When two satellites hit each other, they break into hundreds of pieces. Each piece can smash another satellite. That is a chain with no brake on it.

A new study asks one plain question. How big can a single fleet get before it starts that chain by itself? For many planned fleets, the answer is smaller than the plan.

The test

The paper is called “Critical Sizes of Satellite Constellations”. Hugh G. Lewis wrote it. He works in the Space Environment and Radio Engineering group at the University of Birmingham, in the United Kingdom. He put it on arXiv on 31 July 2026.

Lewis did not launch a thing. He built a maths model instead. It grew out of work by Donald Kessler and Phillip Anz-Meador in 2001, which Lewis and Kessler updated in 2025. Those older models assumed satellites could not dodge. Many modern ones can. So Lewis added dodging. He also added the slow climb up to the work orbit, and the ride back down.

He then ran 16 real fleets through the model. Fourteen of them fill his main table. Between them they hold close to 1.7 million satellites. The biggest is SpaceX Starmind, with one million planned. The smallest is AST SpaceMobile, with 243. He tested every height from 200 km to 2,000 km, in steps of 10 km.

The rules he gave each fleet were kind ones. Each satellite works for five years. Only one in a hundred fails. Dodging works every time, and cuts the crash rate to a hundredth of what it would be. Ninety-nine out of every hundred satellites come down as planned. Those settings match or beat what the best operators do now.

The result

Nine of the 16 fleets are too big. Six of them pass the point where their own wreckage grows without limit. Those six are Guowang, Stampede, Project Sunrise, Starcloud, Starmind, and E-Space Cinnamon and Semaphore counted together. Three more push wreckage up to a high but steady level. They are Eutelsat OneWeb, Eutelsat Next and Starlink Generation 3. The other seven stay under both lines.

Size alone does not decide it. Eutelsat Next plans only 528 satellites and still crosses a line. Starlink MSS plans 15,000 larger ones and stays below both. Height, mass, surface area and dodging all feed into the sum.

What it means

Regulators grade satellites one at a time. A form asks about one craft. It does not ask what happens when the same craft flies in a ring of fifty thousand. This paper gives a way to ask that, using only the numbers a company already files.

The answer comes out as a pass or a fail. That makes it usable in an office, not just a lab. In 2026 alone the United States regulator took in plans for more than 1.2 million satellites. That would grow the number of working satellites by more than 7800 percent.

Small design choices flip the answer. Telesat Lightspeed drops below both lines if its satellites live eight years instead of five. A merged Eutelsat OneWeb and Eutelsat Next would cross the runaway line, but not if the mission runs eight years. Flying under power all the way down to burn-up beat the older five year rule every time.

Business ideas from this paper

Build a web calculator that scores any satellite plan against its critical size, using the equations in this paper. Who buys it: small satellite firms and the law firms that file for them. A price to test: 199 dollars a month per seat. A one-week test: code the model in a spreadsheet, score three public filings, and send the scores to ten space lawyers. Count the replies.
Sell a one-page risk note on a named fleet to the people who insure it. Who buys it: space insurance brokers and underwriters. A price to test: 900 dollars per note. A one-week test: write one note for free on a fleet that failed here, show it to five brokers, and ask what they would pay for ten more.
Run a paid alert list that scores each new satellite filing within 48 hours of it going public. Who buys it: reporters, policy staff and rival operators. A price to test: 29 dollars a month. A one-week test: post four scored filings free on a mailing list, then ask the first 100 readers to pay.

How sure can you be?

This is a model, not a measurement. Nothing was flown and nothing was counted in orbit. The paper has not been through peer review, and it names no journal.

Lewis lists his own soft spots, and they all point one way. Each fleet is judged alone, in an empty sky. Real orbits already hold old satellites, dead rocket stages and debris. He leaves out breakups that are not crashes. He leaves out small bits that could still wreck a satellite. He spreads each fleet through a full shell around the Earth, when real fleets sit in a thinner band. He assumes dodging never fails.

EVERY ONE OF THOSE CHOICES MAKES THE SAFE LIMIT LOOK BIGGER THAN IT IS. In other words, the true limits are likely lower, and more fleets would fail. To settle this, the work needs peer review, a run that includes the debris already up there, and real dodge success rates published by operators.

Do this today

Read the paper before you read the next press release about a new mega fleet. Then ask one question of any plan you see: how big is this fleet, and at what height.

Source: Critical Sizes of Satellite Constellations, July 2026. arXiv:2607.29644 · arxiv.org (preprint · not yet peer reviewed).

Just Out Tech explains new research in plain language. This article was drafted with AI assistance and checked by a human against the original source.

Questions people ask

what is the critical size of a satellite constellation?

It is the largest number of satellites a fleet can hold before its own crashes make wreckage faster than the wreckage clears. Below that number, the fragment count falls over time. Above it, the count grows. Lewis works out this number for each fleet at every height from 200 km to 2,000 km.

which satellite fleets failed the test?

Six fleets crossed the runaway line: Guowang, Stampede, Project Sunrise, Starcloud, Starmind, and E-Space Cinnamon and Semaphore counted as one. Three more crossed the unstable line: Eutelsat OneWeb, Eutelsat Next and Starlink Generation 3. Seven stayed below both lines, including Starlink MSS, Amazon Leo, Lynk, Telesat Lightspeed and AST SpaceMobile.

does this paper prove a Kessler cascade is coming?

No. It is a maths model, not a measurement of orbit. It shows that under the author's rules, nine fleets would each build a growing fragment cloud on their own. The paper has not been peer reviewed and names no journal, so treat it as a warning to check, not a settled result.

why does the paper say its own numbers are too generous?

Lewis judges every fleet in an empty sky, with no old satellites, rocket stages or existing debris. He also assumes collision dodging works every time and cuts the crash rate to one hundredth. He spreads satellites through a whole shell around Earth rather than the thinner band they really fly in. Each choice raises the safe limit, so real limits would be lower.