The old idea
In 1903, a provincial Russian schoolteacher named Konstantin Tsiolkovsky published the equation that still governs every launch: Δv = I_sp · g · ln(m₀ / m_f) The velocity you can gain equals your engine’s efficiency multiplied by the natural logarithm of how much mass you started with divided by how much you finished with. Verne had already fired a crew at the Moon from a cannon in 1865, and Goddard would fly the first liquid-fuelled rocket in 1926. But Tsiolkovsky’s line is the one that mattered, because it converted a dream into an accounting problem with a very unpleasant term in it.
The physics
The unpleasant term is the logarithm. To gain velocity linearly, mass must grow exponentially.
Figure 9.1 — The tyranny Tsiolkovsky described in 1903
Reaching low Earth orbit takes roughly 9.4 km/s once you include drag and gravity losses. Put that into the equation with the best chemical propulsion available and you finish with something like four to eight percent of your liftoff mass and that residue is not payload. It is tanks, engines, structure, avionics and payload. Payload alone is typically two to four percent. Here is the important consequence, and it is why this chapter is short. Nobody is going to improve this. Chemical propulsion is within a few percent of its theoretical energy limits, and has been for decades. There is no propulsion breakthrough coming that changes the cost of getting to orbit. Which means the entire hundredfold cost reduction described in Chapter 2 came from somewhere else: not from making the rocket better, but from not throwing it away. Reuse does not beat the rocket equation. It amortises it. That distinction is worth holding, because it tells you where further gains can and cannot come from, cadence, refurbishment cost and manufacturing volume, not physics.
What breaks
Three things, none of them thermodynamic. Cadence. A reusable vehicle only pays back if it flies often. The economics of the entire orbital compute sector are downstream of how many times per year a heavy vehicle can be turned around, and that is an industrial and regulatory problem, pads, range availability, environmental review, as much as a technical one. Fairing volume, not mass. Chapter 8 showed that a megawatt node is volumetrically modest but mechanically complex. As launch mass gets cheap, the binding constraint moves to how much folded structure you can pack and reliably unfold. Vehicle diameter starts to matter more than vehicle payload. The last mile. A rocket drops you where its trajectory ends, not where your constellation needs to be. Chapter 8’s thermal and radiation analysis pushed us into a narrow altitude band and a specific orbit plane; getting a heavy node into precisely that plane, and later moving or disposing of it, is a separate vehicle and a separate business.
Who is attacking it
SpaceX is the layer, in a way that has no parallel elsewhere in this book. It sets the price everyone else models against, it has disclosed the only Starship contract price in the public record, and it is simultaneously a competitor in orbital compute, having filed for a very large constellation of compute satellites and merged with an AI company. Any honest treatment of this sector has to describe SpaceX as both the enabling infrastructure and the largest single competitive risk to everyone using it. Rocket Lab and Stoke Space matter as the second sources. A sector whose economics rest on one provider’s price list is a sector with a single point of commercial failure, and the value of a credible second heavy vehicle is greater
than its market share suggests. Impulse Space is the last-mile answer: orbital transfer vehicles that take a payload from where the rocket left it to where it needs to be, in hours rather than months of natural drift. Founded by SpaceX’s first employee, revenuegenerating, with flight heritage and a large backlog.12
The investment stance, stated plainly
I do not think launch is where the returns are, and I do not hold it. Not because it will do badly, it may do very well but because the largest player is about to be, or already is, a publicly discoverable asset, and the second tier is public too. Whatever excess return existed in launch has been competed into the open. The last mile is different. It is a smaller market, it is not price-discovered, and it becomes structurally more valuable exactly as the thermal and radiation constraints of Chapters 5 and 6 push compute constellations into specific, awkward orbits. Value in this layer has moved from getting to space to getting somewhere specific in space.
What to watch
Starship flying paying customers at or near the contracted price, rather than the target price. Turnaround time between heavy flights, the number that determines whether $600/kg becomes $200/kg. Whether a second provider reaches heavy-lift reuse. Until one does, every model in this book has a singlesupplier risk sitting under it.