Sp Space-based data centers Loading... : Investor Sentiment and Bull/Bear Views

Space-based data centers will soon be cheaper than terrestrial ones because orbital solar generates eight times more energy per square meter, avoids permitted land and battery storage costs, and provides continuous 24/7 power in dawn-dusk sun-synchronous orbit. Launch cost break-even is around $500/kg, within reach of Starship's projected $10-$20/kg. Starcloud plans an 88,000-satellite constellation at about 200 kW each, enabling roughly 20 GW of inference capacity for about $100B capex, less than terrestrial equivalent, making it the largest infrastructure project ever.

Johnston argues that space will become the primary location for AI compute within the next decade because launch costs are falling while terrestrial data-center constraints rise. Space avoids permitting and land costs, needs no battery backup because satellites stay in sunlight, uses about eight times less solar panel area, and can dissipate heat via high-temperature radiators. He expects close to $1 trillion per year of CapEx deployed in space within 10 years and at least half of all new compute capacity to be in space within 5-10 years, mostly for inference workloads with Starlink-like sub-50ms latency.

StarCloud's first satellite demonstrates that AI inference and training can run in orbit, and Johnston argues compute should move to space over the medium-to-long term because space offers nearly unlimited low-cost solar energy. He estimates all-in energy costs around an order of magnitude cheaper than new North American energy projects, below 1 cent per kWh versus 5-10 cents, and says hundreds of gigawatts per year can be launched, enabling very large orbital data centers such as a planned 5 GW facility that would be difficult to power on Earth.