The Data Centers in LEO

Where Part II leaves us

ConstraintStatusWho has to solve itPowerPhysics strongly favourable, roughly 8×Array and power-management suppliersper square metre. Engineering is(Ch 10)deployment, high-voltage handling and degradation. HeatSolved below ~500 kW with decades ofThermal specialists (Ch 11), the thinnest

heritage. Unsolved at megawatt scale.

layer in the sectorThe binding constraint. RadiationMaterially de-risked for dose by Google’sCompute and power-electronics designerstesting. Open on long-duration and single-(Ch 12)event behaviour. A depreciation schedule, not a wall. BandwidthNot a constraint for the natural firstOptical terminals and relay operators (Chworkloads. A hard constraint for uploading

14)

terrestrial data.

One of those four rows is materially worse than the others, and it is the one with the fewest investable companies attacking it. That asymmetry is the most actionable observation in this book, and Part III now takes it apart component by component, before Part IV names the people trying.


Download as PDF