In brief
\nThe FCC accepted SpaceX\u2019s application to build a satellite system designed as orbital data centers and opened it for public comment.
\nThe proposal would link laser-connected satellites with SpaceX\u2019s existing Starlink constellations rather than provide consumer broadband.
\nThe filing follows FCC approval of Starlink\u2019s expanded Gen2 network and Musk\u2019s push to shift AI computing off Earth.<\/p>\n
SpaceX CEO Elon Musk\u2019s dream of turning the night sky into a massive, solar-powered brain for artificial intelligence moved a step closer to reality this week as federal regulators began reviewing a plan to launch the company’s new satellite cluster.<\/p>\n
The Federal Communications Commission on Wednesday opened public review<\/a> of a SpaceX proposal to build a non-geostationary satellite system that would move energy-intensive AI computing into orbit, potentially allowing the company to deploy up to one million data-center-style satellites to train xAI models, including Grok.<\/p>\n \u201cThe proposed satellites will use high-bandwidth optical inter-satellite links and conduct telemetry, tracking, and command (TT&C) operations,\u201d the FCC wrote. \u201cThe Bureau seeks comment on the application and the associated requests for waiver.\u201d<\/p>\n \ufeff<\/p>\n The filing review follows Musk\u2019s decision\u00a0on Monday to\u00a0fold<\/a> his artificial intelligence startup, xAI, into SpaceX, consolidating AI development and launch capabilities within a single company.<\/p>\n \u201cThe SpaceX Orbital Data Center system will allow SpaceX to begin delivering much-needed energy-efficient AI compute for consumers, enterprises, and government users around the world,\u201d SpaceX wrote<\/a> in a waiver request for the filing.<\/p>\n Under the proposal, SpaceX would operate the satellite system at altitudes between about 310 and 1,240 miles, linked through laser-based optical connections.<\/p>\n The network would connect with SpaceX\u2019s existing Starlink constellations, allowing data to be routed and processed in orbit before transmission to ground stations.<\/p>\n AI Data centers are quickly becoming one of the biggest new sources of electricity demand<\/a> as AI systems scale. In the U.S., they used about 183 terawatt hours of power in 2024, roughly equivalent to the annual energy consumption of Pakistan. That figure is expected to climb as AI training and usage expand.<\/p>\n In its application, SpaceX described the project as a step toward becoming a \u201cKardashev<\/a> II-level civilization,\u201d a theoretical measure of a society capable of harnessing the full energy output of its star.<\/p>\n \u201cGlobal electricity demand for AI simply cannot be met with terrestrial solutions, even in the near term, without imposing hardship on communities and the environment,\u201d Musk said in a statement<\/a>. \u201cIn the long term, space-based AI is obviously the only way to scale. To harness even a millionth of our Sun\u2019s energy would require over a million times more energy than our civilization currently uses.\u201d<\/p>\n The non-geostationary orbit system would also represent a departure from the company\u2019s consumer-focused Starlink internet service, instead positioning satellites as space-based computing infrastructure designed to operate beyond the power and cooling constraints of AI development on the ground.<\/p>\n