The old idea
In 1978 a NASA scientist named Donald Kessler published a paper describing what happens when the density of objects in orbit passes a threshold: collisions generate fragments, fragments cause further collisions, and the process becomes self-sustaining. The Kessler syndrome is now the single most cited constraint on the long-term use of low Earth orbit, and it was described before the first Shuttle flight.
The physics
Orbital velocity is about 7.5 km/s, and two objects can meet at up to twice that. Kinetic energy goes as the square of velocity, which produces numbers that do not match intuition at all.
Why compute nodes are different, and worse
Here is the point this chapter exists to make, and I have not seen it made elsewhere. An orbital data centre is, per tonne, the largest target ever flown. Chapter 8’s reference node carries roughly 290 m² of solar array and 140 m² of radiator, over 400 square metres of cross-section on a 4.7 tonne vehicle. A conventional satellite of similar mass presents a few square metres. The collision cross-section is up to a hundred times higher for the same mass in orbit.
Figure 16.1 — A compute node is the largest target ever flown
Run the flux against that area over five years and the picture separates cleanly into three regimes: Sub-millimetre impacts are constant. Thousands of them. They erode coatings, which attacks emissivity, which sits inside Chapter 5’s fourth-power law. This is a slow degradation of thermal capacity, not an event. Millimetre-class impacts are certain, on the order of twenty over five years for one node. Most hit array or structure and do little. Some hit a radiator, and a radiator is a thin, fluid-filled surface. This is why loop isolation and panel redundancy are requirements rather than refinements, and it is the specific reason the ISS radiators were designed with debris protection built in. Centimetre-class impacts are a low-probability, total-loss event, a couple of percent per node over five years on these numbers. For a single satellite that is an acceptable risk. For a constellation of a thousand nodes it means you should expect to lose roughly twenty of them to debris, and you must have designed for it.
The constraint nobody is modelling: insurance capacity
The space insurance market is small. Global annual premiums are on the order of half a billion dollars, less than the cost of a single large terrestrial data centre campus and the market has had loss-making years recently.17 Now consider what this sector proposes to insure. A gigawatt-class orbital constellation is tens of billions of dollars of hardware, in an environment where the loss modelling is immature, the failure modes are correlated across the fleet, and the largest single asset class has no actuarial history whatsoever. The binding constraint on scaling this sector may turn out to be underwriting capacity rather than physics. Debt financing requires insurance. Insurance requires a loss model. A loss model requires history. The sector has none, which means early constellations will either be equity-financed at high cost of capital, self-insured by balance sheets large enough to absorb the loss, or state-backed. All three of those favour very large incumbents over startups, and that is a competitive dynamic worth understanding before it becomes obvious.
Who is attacking it
Space situational awareness providers, LeoLabs, Slingshot and the government catalogues, sell the tracking that makes conjunction avoidance possible. Specialist brokers and underwriters price the risk. Debris removal companies exist and remain in search of a customer willing to pay.
The investment stance
I do not hold this layer, and I want to be clear that this is a sizing judgement rather than a quality judgement. Tracking and insurance are real businesses; they are simply too small, at current scale, to move a concentrated
fund, and debris removal has no proven buyer. But this chapter belongs in a guidebook because it is a cost line and a gating risk in every other chapter. Every operator’s model needs a debris allowance, a conjunction-avoidance propellant budget, and an insurance premium and most published models contain none of the three.
What to watch
The first insurance policy written on an orbital data centre, and its rate. That number is the market’s honest estimate of the risk, and it is more informative than any technical paper. Any conjunction event involving a very large-area spacecraft. The first one will reprice the entire sector’s risk assumptions in a week. Whether regulators begin treating cross-sectional area, rather than satellite count, as the quantity to limit. That would change constellation design more than any other plausible rule.