Why the economics of orbital AI are so brutal

The article discusses the emerging concept of orbital AI data centers, driven by Elon Musk and other tech leaders envisioning vast networks of satellites hosting AI compute power in space. SpaceX has proposed solar-powered orbital data centers distributed across up to a million satellites, potentially delivering 100 GW of compute power off-planet. Other players, including Google with its Project Suncatcher and startups like Starcloud, are also pursuing similar ambitions. Despite the hype, current analyses show that terrestrial data centers remain significantly cheaper than orbital alternatives. For example, a 1 GW orbital data center could cost around $42.4 billion, nearly three times the cost of an equivalent ground-based facility, mainly due to the high upfront expenses of satellite manufacturing and launch. A critical factor in making space-based data centers economically viable is drastically reducing launch costs. While SpaceX’s Falcon 9 currently costs about $3,600 per kilogram to orbit, Project Suncatcher estimates that prices need to fall to around $200/kg—an