The old idea
In 1929 an Austro-Hungarian army officer writing under the name Hermann Noordung published The Problem of Space Travel, containing the first detailed engineering description of a habitable rotating space station, a structure obviously too large to launch whole, and therefore obviously something you would assemble in orbit. Von Braun popularised the same vision in Collier’s in 1952. O’Neill industrialised it in 1976. Assembly in orbit is therefore among the oldest ideas in spaceflight, and unlike most of the ideas in this book it has actually been done. Skylab was saved in 1973 by astronauts deploying a sunshade over a station that had lost its thermal shield. The Hubble Space Telescope was repaired in 1993 and serviced four more times. The International Space Station was built from dozens of separately launched modules. So the question is not can we assemble and service things in orbit. It is whether it can be done without astronauts, and at a price that a commercial operator would pay. Every servicing success so far has been a national programme costing billions.
The physics, and the constraint that decides everything
Getting from one satellite to another is not like driving to another building. Orbits are defined by six parameters, and changing some of them is nearly free while changing others is ruinous.
Figure 15.1 — Servicing only works inside one orbital plane
Moving to a different position within the same orbital plane costs almost nothing: you change altitude slightly, drift ahead or behind, and circularise. Tens of metres per second, plus patience.
Changing the plane costs Δv = 2v·sin(Δi/2). At an orbital velocity of 7.5 km/s, one single degree costs about 132 m/s, more than raising your altitude by two hundred kilometres. Ten degrees costs over 1,300 m/s, which is most of a servicer’s entire propellant life for one trip. From that one relationship comes the most actionable design rule in this chapter: A constellation that intends to be serviced must be co-planar from its first launch. Serviceability is not a capability you add later; it is an orbital architecture decision made before anything flies.
There is one loophole worth knowing. Earth’s equatorial bulge causes orbital planes to precess, and the rate depends on altitude and inclination. A servicer parked at a slightly different altitude drifts in plane angle for free, at perhaps a degree every few days. Time substitutes for propellant. That works beautifully for a patient logistics vehicle and not at all for a repair call.
Why this chapter matters more than its market size suggests
Chapter 8 ended on an unsolved problem: spacecraft last five to fifteen years, AI accelerators are economically obsolete in about three, and an orbital operator currently has no way to roll in new racks. This is the chapter where that gets solved, or does not.
The counter-argument is Chapter 13’s: if modules are cheap enough, you do not service them, you replace them whole and deorbit the old ones. Which answer wins depends on the ratio between the payload’s value and the rest of the vehicle’s value and on Figure 8.5’s unresolved question of what the power hardware actually costs. If arrays and radiators turn out to be cheap, disposable modules win and this chapter is a footnote. If they are expensive, servicing is the sector’s most valuable unbuilt business.
What breaks
Autonomous rendezvous and proximity operations without a cooperative target. Docking interfaces, which do not exist as an industry standard for this application. Thruster plume impingement on the delicate, enormous surfaces of Chapter 8. Fluid transfer for two-phase thermal loops, which is materially harder than propellant transfer. And the fact that a servicer approaching a $50M asset can destroy it.
Who is attacking it
Impulse Space is the mobility layer described in Chapter 9, and mobility is the prerequisite for logistics. Outpost is working on Earth return, which is the other half of the loop, bringing hardware down rather than only up. Starfish Space, Astroscale and the established primes’ servicing vehicles have demonstrated approach and docking with uncooperative targets. Europe’s robotic assembly demonstrator programme is flying its first mission in this window.16
The investment stance
This layer has disappointed investors for fifteen years, and I hold it lightly and specifically. The pattern has been consistent: technically impressive demonstrations, and a customer base that turns out to prefer buying a new satellite. What is different now and the reason to keep watching rather than dismiss it is that the customer has changed. Previous servicing markets were about extending the life of a satellite whose whole value was ageing. Orbital compute is the first application where a small, valuable, rapidly-obsoleting component sits inside a large, slowly-obsoleting vehicle. That is precisely the structure that makes servicing economic.
So: not a bet on servicing as a category. A bet on mobility, with servicing as the option attached.
What to watch
The first commercial payload swap on orbit, in any application. It would reprice this layer immediately. An orbital compute operator publishing a docking interface standard, which would tell you it has decided serviceability is worth designing for. Whether the compute constellations now being filed are co-planar. Figure 15.1 means that single architectural choice determines whether servicing is even possible for them.