The old idea
This is the layer with the deepest intellectual pedigree, and it is worth spending a paragraph on because it reframes what the sector is doing. In 1964 the Soviet astronomer Nikolai Kardashev proposed classifying civilisations not by their politics or their technology but by the energy they command: a Type I civilisation uses the energy available on its planet, a Type II the output of its star. The scale was proposed as a tool for searching for extraterrestrial intelligence, but its underlying claim is the one that matters here, that the ceiling on what a civilisation can do is set by the power it can gather, and that the next available increment is not on the planet. Four years later, in 1968, the engineer Peter Glaser published the solar power satellite concept: collect sunlight in orbit where it is uninterrupted, and send the energy down. Gerard O’Neill built an entire architecture on this premise in The High Frontier in 1976. Asimov had already written the fictional version in 1941. None of it happened, for one reason. At Space Shuttle prices, putting a kilogram in orbit cost roughly fifty thousand dollars, so the energy hardware could never repay its ride. The idea was never wrong. It was priced out. Chapter 2 is the story of that price changing, and this chapter is the first place where the change bites.
The physics
Four ways to have a watt in orbit, and they differ on axes that matter more than efficiency.
Figure 10.1 — Four ways to get a watt in orbit, and what each one costs you
Photovoltaic is the default: high specific power, no fuel, but useless in eclipse and dependent on pointing and deployment. Within it sits the multi-junction versus silicon argument from Chapter 8, efficiency per square metre against cost per watt at manufacturing scale. Beamed power is the interesting one, and it deserves more attention than it gets. The insight is not that beaming sunlight is more efficient, it is not, since you pay conversion losses twice. The insight is who owns the receiver. If you concentrate and beam sunlight onto a client spacecraft’s existing solar array, the customer needs no new hardware at all: no retrofit, no custom receiver, no integration programme. That converts a hard sale into an easy one, and it turns a power company into shared infrastructure serving every operator rather than a supplier picking a winner. Radioisotope power is the opposite trade: a few watts per kilogram, which is terrible, in exchange for output that does not care about eclipse, sun angle, distance from the Sun, or dust. It will never power a megawatt data centre. It is the correct answer for anything that must survive the dark. Fission is the only technology that could plausibly deliver megawatts without square kilometres of array and the thermal analysis of Chapter 5 applies to it doubly, since a reactor’s waste heat must also be radiated. There is no flight heritage at the relevant scale.
What breaks
Everything in Chapter 8’s power section, plus the number that chapter refused to let pass: cost per watt. Recall
Figure 8.5 — historical bespoke space arrays on the order of $1,000/W against an implicit assumption near $3/W in
gigawatt-scale plans. That gap is this layer’s central commercial question, and any company here that cannot answer it precisely is selling a story. Beyond it: deployment mechanisms, high-voltage arcing in LEO plasma, radiation degradation of cells, and, for beamed power, pointing accuracy and conversion efficiency at range.
Who is attacking it
K2 Space is the clearest structural bet in the layer, because it does not sell power as an accessory to a satellite; the satellite is the power system. A large bus designed from the start around high power, high-voltage avionics and a high-power thruster, sized for what heavy lift now permits rather than for what a Falcon-class fairing used to permit. It has raised at a multi-billion valuation against substantial signed contracts and has won a programme of
record.12 Rocket Lab entered this layer in February 2026 with silicon arrays explicitly aimed at gigawatt-scale space data centres, betting on manufacturability over efficiency.10 Star Catcher is the beamed-power bet described above, concentrating and beaming solar energy onto clients’ existing arrays, with signed power purchase agreements ahead of first flight and multiple orbital-compute companies among its named customers.12 Its founder previously built and sold an in-space manufacturing company, which is the repeat-founder profile that historically predicts more than any other single signal. Zeno Power is the radioisotope bet, and its moat is unusual: not physics, radioisotope power is sixty years old but a fuel supply chain. Recycled strontium-90 and americium-241 with contracted industrial partners is not something a competitor replicates by raising money faster.12 Aetherflux sits across two layers, generating power and consuming it in its own compute payload. Strong founder, top-tier syndicate, and, at the time of writing, nothing yet demonstrated in orbit, which is the correct reason to hold it smaller than its narrative would justify.
The investment stance
Power is where I am most concentrated, and the reason is structural rather than enthusiastic. Every operator in this sector needs watts, and none of them agree on whose compute wins. A supplier that sells to all of them does not require me to pick the winner in the layer above, which matters enormously over a five-year hold through a period when the operator layer will consolidate. Chapter 8 also showed that power hardware carries the single largest unresolved cost in the whole thesis, and unresolved costs are where pricing power lives. The caution: this layer’s valuations are the most sensitive to the $/W question. If mass-produced arrays arrive at $3/W, the power layer becomes a commodity and the margin moves elsewhere. If they arrive at $100/W, the sector’s gigawatt plans do not happen and the layer is small. Both outcomes are bad for a naive long position, which is why the specific companies here are chosen for moats, a supply chain, a shared-infrastructure position, a bus architecture, rather than for exposure to the theme.
What to watch
The first published dollars-per-watt for a mass-produced orbital array. This single number moves more of the model than anything else in this book. A successful spacecraft-to-spacecraft power beaming demonstration on orbit. It has never been done commercially; doing it creates a new category. Whether high-power buses win programmes of record against high-rate manufacturers. That contest tells you whether power or production volume is the durable differentiator.