SpaceX Files for 1 Million Satellite Orbital Data Center Constellation

SpaceX filed the application three days before its IPO while Elon Musk stated at Davos that orbital AI would become lowest-cost within two to three years. The proposal follows a single Nvidia H100 test flight by Starcloud that required derated power due to insufficient radiator area of 1.4 square meters for 700 W dissipation. Scaling to a 100 MW facility demands 2,500 separate 80 m² radiators, exceeding current Starlink manufacturing rates of 4,000 satellites per year by two orders of magnitude.

Current orbital launch records show 165 SpaceX missions in 2025 and 7,000 total historical launches. Deploying 1 million satellites at 60 per Starship requires 16,666 dedicated flights, projecting a minimum decade at tenfold cadence increase. Radiative cooling limits and Kessler syndrome collision risk remain unaddressed in public filings.

The coverage understates strategic drivers: orbital placement bypasses terrestrial grid constraints projected at 8% of U.S. electricity demand by 2030 and enables sub-20 ms latency to polar and maritime regions outside fiber routes. Control of spectrum and orbital slots also confers de facto compute sovereignty advantages in contested spectrum bands.

Operational deployment will be gated by radiator mass fraction and on-orbit servicing cadence rather than launch cadence alone. No public thermal or power budget for multi-GPU racks has been released beyond the single H100 test article.