Particle.news
Download on the App Store

Companies Propose Massive Orbital AI Data‑Center Fleets as New Analyses Question Feasibility

Recent FCC filings and startup roadmaps promise tens of thousands to a million compute satellites, with engineering assessments finding cooling, launch and manufacturing limits that make mass deployment unlikely in the near term.

Overview

  • Multiple firms have filed formal FCC applications this month proposing very large constellations: SpaceX’s filing seeks permission for up to 1 million compute satellites and Orbital has asked to deploy up to 100,000 satellites with a 10‑gigawatt target.
  • Orbital’s filing describes 100‑kilowatt‑class satellites with roughly 100‑meter solar arrays and radiator surfaces, plans for a small demonstration next year, and a first operational Orbital‑1 satellite targeted for 2028.
  • Independent technical analyses published July 1 calculate severe thermal limits for space cooling, showing a single Nvidia H100 GPU needs about 1.4 m² of radiator and that rack‑scale and data‑center‑scale power would demand impractically large radiator areas.
  • Analysts also flag practical bottlenecks in launch cadence and manufacturing scale, noting that lofting hundreds of thousands to millions of satellites would require launch and production rates far above historical levels and heavy reliance on vehicles like Starship.
  • Regulators, astronomers and some investors have raised concerns about approvals, orbital traffic and debris, and the economic case for orbital compute, so most experts expect stepwise demonstrations and slow scaling rather than immediate mass rollouts.