Orbital data centers could address the growing difficulty of powering terrestrial AI infrastructure, where grid connections and gas-turbine supplies face long delays. Their main advantage is energy: solar arrays in dawn-dusk orbit receive near-continuous sunlight and can generate three to five times as much power as comparable panels at good Earth sites. But current launch costs make space-based power uneconomic. The analysis estimates that space solar could match present-day off-grid terrestrial power at about $250 per kilogram to orbit, and become cheaper than today’s terrestrial energy options at roughly $50 per kilogram. Reaching those prices at the scale required depends chiefly on Starship becoming reusable and operating at high cadence; it is the only near-term launch system judged capable of enabling substantial growth.
Cooling, often treated as a decisive obstacle, appears manageable: lightweight radiators could make thermal systems a modest share of costs, potentially below terrestrial cooling expenses. More serious constraints include the inability to repair or replace failed chips, which could require roughly 40% extra launch and supporting hardware over a facility’s lifetime, as well as limited links between satellites that may initially restrict orbital systems to inference. Radiation, supply-chain immaturity, orbital congestion, regulation, and security also remain unresolved concerns. The assessment is that orbital facilities are unlikely to supply a meaningful share of compute before 2030, but could become cost-competitive within three to five years if Starship progresses as planned, potentially supporting a substantial portion of data-center expansion during the 2030s. Their growth could also shift AI governance and concentrate influence around a small number of launch providers.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.