what is a critical size?

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

who would pay for a note like this?

Space underwriters and brokers who must price a fleet they have never seen, and small operators whose insurer has started asking. In 2026 alone the United States regulator took in plans for more than 1.2 million satellites, so the reading load is the problem you are selling into.

what should the note never say?

That a fleet is safe. The model judges each fleet alone in an empty sky, assumes dodging always works and spreads satellites through a whole shell. Every one of those choices makes the safe limit look bigger than it is, so a pass is a screening result and not a verdict.

how strong is the evidence?

We score it 3 out of 10. It is an unreviewed preprint with no code, and it is a maths model rather than a measurement of orbit. It is lifted by a single author with a long record on this exact model, applied to 16 real fleets.

Sell insurers a one-page orbit crowding score

The finding this rests on

A researcher at the University of Birmingham worked out the largest number of satellites a single fleet can hold before its own crashes make wreckage faster than the wreckage clears. He calls that the critical size. Then he ran 16 real fleets through it.

Nine of the 16 sit above their own critical size. Six of those nine cross the worst line, where fragments grow without limit. Those six are Guowang, Stampede, Project Sunrise, Starcloud, Starmind, and E-Space Cinnamon and Semaphore counted together. Three more sit above the first line: Eutelsat OneWeb, Eutelsat Next and Starlink Generation 3. The full article is at nine of 16 planned satellite fleets are too big for orbit.

Size alone does not decide it. Eutelsat Next plans 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.

We score this paper 3 out of 10. That is low, and you should treat this idea as a bet on the finding holding up. It is a preprint with no journal named and no code released. It is a maths model, so nothing was flown and nothing was counted in orbit. What lifts it above the floor is the author, who has a long record on this exact model, and the fact that he applied it to 16 real fleets.

Who has this problem

Picture a four person space underwriting team at an insurance market in London. They price cover for satellites and launches. A broker sends over a fleet nobody in the room has heard of, and asks for terms by Friday.

The filing is long. It gives the satellite count, the height, the mass and the plan for bringing craft down at end of life. All of that is public. None of it answers the question the team actually needs answered. Does this fleet make the orbit worse for everything else we already cover?

The volume is the hard part. 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. Nobody has time to model each one by hand.

A small operator has the mirror of this problem. They plan 60 craft at one height. Their insurer asks a question they have never had to answer, and they have no way to show their answer is reasonable.

What you would sell them

Sell a one-page note on one named fleet. The top of the page says whether the fleet sits above or below its own critical size at its planned height, and by how far. The rest of the page shows the levers.

The levers are the useful part. In the paper, Telesat Lightspeed drops below both lines if its satellites live eight years instead of five. Flying under power all the way down beat the older five year disposal rule every time. So your note tells an underwriter which change would move this fleet across the line, and tells an operator which change would move their own.

On Monday morning the underwriter has one page in front of the broker, with a number, a height and three levers. That is what they take into the pricing meeting.

What to charge

Test it at 900 dollars a note. That is a hypothesis, and you should say so on the first invoice you send.

Set it against the hours the team already spends. Two of their own people reading one filing for a day is the thing you replace. You know what a day of your own analyst time costs. Put your number beside 900 dollars and see which way it falls.

Then test a second shape. A retainer, at a monthly price, that covers every new filing plus two full notes. Underwriters buy the same thing over and over, so a per-note price may be the wrong unit. Find out which one they reach for.

How you would build the first version

Build it in a spreadsheet. Put the model’s equations in one sheet and the fleet inputs in another. The inputs are all public. Satellite count, height, mass, surface area, mission length and the disposal plan come from filings, and the paper itself cites public sources such as ESA DISCOS.

Score across the whole height range, from 200 km to 2,000 km, the way the paper does. A fleet that passes at its planned height may fail 100 km lower, and that is worth a line on the page.

Do not build a web app. Do not build a live tracking feed. Do not build an interface for operators to run themselves, because then you have sold a tool instead of a judgement, and the judgement is the thing they cannot copy.

Write the first three notes by hand, in a text file, and read them out loud. If a note cannot be understood by someone who has never read the paper, it is not finished.

The one-week test

Day one. Put the model in a spreadsheet. Check it by reproducing the paper’s own result for three of the 16 fleets. If your numbers do not match the paper, stop and fix the sheet.
Day two. Pick a fleet the paper did not cover. Pull its public filing. Score it. Write the one-page note.
Day three. Find 15 people who price or broker space risk, and 10 small operators with a fleet in planning. Get names, not company inboxes.
Day four. Send the free note to all 25 with one question. Would you pay 900 dollars for this on a fleet of your choosing, and we are testing that price. Ask for a named fleet and a purchase order.
Day five. Count purchase orders. Two paid notes from 25 is a yes. Zero, with polite replies saying it is interesting, is a no.

What would kill this

The model is one person’s preprint. It has not been through peer review and names no journal. An underwriter who takes your note to a committee may be told exactly that. Put the score on the page yourself, before someone else does.

A pass may be wrong in the dangerous direction. The author judges each fleet alone in an empty sky, with no old satellites, no dead rocket stages and no existing debris. He assumes dodging works every time and cuts the crash rate to a hundredth. He spreads satellites through a whole shell rather than the thinner band they really fly in. Every one of those choices makes the safe limit look bigger. A PASS IN THIS MODEL IS NOT SAFETY. Say that on every page.

You may be selling a verdict you cannot stand behind. If an insurer prices a risk on your note and the note is wrong, the argument lands on you. Sell it as a screening question that tells them where to look, and write that limit into the contract.

Buyers may already have this covered. Larger insurers keep their own engineers. Some operators have modelled their own fleet for years. Day four exists to find out whether anyone outside that group will pay.

The research may not hold. Nothing was flown and nothing was measured. What would settle it is peer review, a run of the model that includes the satellites and debris already in orbit, and real dodge success rates published by operators. Until then, every note you sell rests on a model, and your buyer deserves to be told so in the first line.

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.