Overview
- A joint announcement Tuesday, August 4, said SpaceX and NVIDIA will co‑develop the Starmind AI1 satellite payload using NVIDIA’s Space‑1 Vera Rubin NVL72 hardware to process advanced AI workloads in low Earth orbit.
- Elon Musk confirmed SpaceX will use NVIDIA GPUs exclusively and the companies target rapid scale of compute capacity with company guidance pointing to multi‑gigawatt deployments by 2027.
- Public descriptions give each Starmind AI1 satellite roughly 120 kW average compute and up to 150 kW peak, with large solar arrays and Starlink optical (laser) links planned to move data between satellites and back to Earth.
- The program still needs final FCC authorization for SpaceX’s earlier filing to deploy up to one million data‑center satellites, and independent engineers warn of major hurdles including heat rejection in vacuum, chip radiation damage, launch cadence and collision and debris risks.
- Prototype testing is slated for early 2027 with mass production aimed for late 2027 if development and regulatory reviews proceed, a timetable that could reshape NVIDIA’s market demand and raise new questions for astronomers, launch crews and satellite operators.