Putting computers in space is not a new idea. Nothing in this book is a new idea; that is rather the point, and each sector chapter in Part IV opens by naming the person who had it first. What changed is not the concept. What changed is a price.
Figure 2.1 — Two orders of magnitude in one working lifetime
The Space Shuttle delivered payload to low Earth orbit at something on the order of fifty thousand dollars per kilogram. A dedicated Falcon 9 today lists around two thousand seven hundred to three thousand. And in March 2026, for the first time, a Starship price entered the public record not through a press release but through an SEC filing. Voyager Technologies disclosed a $90 million contract for a single launch, and Voyager’s payload is the Starlab station, which is designed to go up whole rather than be assembled in pieces.6 Divide the disclosed contract by a station mass in the hundred-to-one-hundred-fifty-tonne range and you get something near six hundred dollars per kilogram.7 Be careful with that number, and be careful with it in public. It is an inference, not a published price: the filing gives a contract value, and the mass comes from the station’s design, and nobody at either company has confirmed the division. It is also a price for a single very large dedicated payload, which is the most favourable case there is. And Starship has not yet flown that mission. Treat $600/kg as the best evidence currently available rather than as an achieved fact, and treat the company’s own long-run guidance of $100, 200/kg as what it is: a target.
Why reuse changes the shape of the curve, not just its level
A first-principles reader should not accept “reuse makes it cheaper” without asking how much cheaper, and why there is a floor. The cost of a launch decomposes into four things: propellant, hardware, operations, and the amortisation of the development programme. Propellant is startlingly cheap, the liquid oxygen and methane in a large vehicle costs on the order of a million dollars, which is a rounding error against a $90 million contract. Hardware is expensive and, in an expendable rocket, is destroyed on every flight. Operations and refurbishment sit in between.
So the arithmetic of reuse is simply this: if the hardware costs H to build and flies N times, its contribution per flight is H/N plus refurbishment. Going from N = 1 to N = 10 does not make launch ten percent cheaper; it removes ninety percent of the largest line item. Going from N = 10 to N = 100 does much less, because by then propellant, range operations and refurbishment dominate. That is why the curve in Figure 2.1 falls steeply and then flattens, and it is why $100/kg is difficult in a way that $600/kg is not.
The ratio that decides everything
Here is the calculation that tells you whether any of this is serious. It takes four inputs and you should change all of them as you read.
That is the entire reason this book exists in 2026 and could not have been written in 2016. The physics of orbital compute was as sound then as it is now. It was simply unaffordable, and unaffordable ideas look identical to wrong ideas until the price moves. One more consequence, and it is the one most people miss. When launch is the dominant cost, every engineering decision optimises for mass. When launch becomes cheap, mass stops being the master constraint and the design is free to get heavy where heaviness helps, thicker shielding, bigger radiators, deployable structures with margin. Much of Part IV is about companies who have understood that the design rules changed and are building for the new regime rather than the old one.