Industry leaders question the technical and economic viability of Elon Musk's orbital data center proposals
Moving heavy compute into low Earth orbit severely underestimates the physics of thermal management in a vacuum, where radiative cooling is the only option. While solar power is abundant, the mass-to-orbit costs for radiation-hardened hardware and latency penalties make this a highly inefficient architecture for standard AI workloads.
What Happened
SoftBank CEO Masayoshi Son and other industry executives are publicly questioning Elon Musk's recent propositions regarding "orbital data centers." Musk has floated the idea of leveraging SpaceX's heavy-lift capabilities to place massive compute clusters in space, theoretically bypassing terrestrial power grid constraints and benefiting from continuous solar energy. However, the broader tech and telecommunications industry is pushing back on the feasibility of this vision.Technical Details
From an engineering perspective, orbital compute faces brutal physics constraints, primarily around thermodynamics. While space is cold, it is also a vacuum. Terrestrial data centers rely heavily on convective cooling (air or liquid) to manage the massive heat output of modern AI accelerators. In orbit, the only mechanism to shed heat from 700W+ GPUs is radiative cooling, which requires massive, heavy radiator panels.Additionally, cosmic radiation necessitates either heavy physical shielding or specialized, slower radiation-hardened silicon—both of which destroy performance-per-watt and performance-per-kilogram metrics. Finally, while Low Earth Orbit (LEO) offers lower latency than traditional GEO satellites, the round-trip data transmission still adds a hard physics floor of 10-30ms. Combined with atmospheric interference and routing overhead, this makes orbital compute unsuitable for real-time inference or high-bandwidth training clusters.