The Orbital Data Center Hype Machine Is Already in Orbit
TL;DR
IEEE Spectrum punctures the orbital data center narrative: SpaceX reportedly filed with the FCC for a low-Earth-orbit constellation of up to 1 million satellites. The scale problem is severe. There are about 14,500 active satellites today; deploying 1 million units on Starship would require roughly 16,666 dedicated launch missions.
Nauti's Take
This is a classic AI infrastructure narrative: a real bottleneck gets wrapped in a spectacular solution that is still far from operational reality. Space-based data centers may eventually serve niche workloads, especially for a company that controls launch, energy, and network infrastructure.
As an answer to today’s data center crunch, though, it reads more like an investor story than an operator plan. The sober benchmark is simple: cool one H100 reliably before promising a million satellites.
Briefingshow
Orbital compute sounds like a neat answer to power limits, land constraints, and exploding AI demand. But the economics weaken once launch cadence, manufacturing, cooling, maintenance, latency, and orbital debris are counted together. The key issue is not whether space compute is imaginable, but who benefits from selling that timeline now.
Sources