Op-Ed: Don't Believe the Hype; Space Data Centers Are Way Off

Beaming our computing problems into orbit sounds futuristic; the reality, argues one expert, is mostly hype and hidden costs. Writing in the Financial Times, Rebekah Reed of Harvard's Belfer Center dissects SpaceX's recent pitch to build a "constellation of a million satellites" serving as orbital data centers—a vision Elon Musk has floated as just a few years away, and one that appeals to tech giants grappling with AI's ballooning energy use. Reed says space-based computing makes sense in narrow cases, like processing data already generated in orbit. But as a way to dodge AI's power and cooling demands on Earth? She calls that timeline wildly optimistic.

To make it feasible, launch costs would need to plunge to roughly $200 per kilo, which is not expected until the mid-2030s. This would also require specialized space-hardened hardware that doesn't yet exist at scale. Add in the complicated issue of maintenance becoming a high-stakes space mission, potential emissions rivaling or exceeding those of ground data centers, atmospheric impacts from burning up old satellites, and growing risks from orbital crowding, and the "quick fix to maintain AI's growth curve" looks anything but. Click for the full analysis, including Reed's suggestions for how to improve the efficiency of data centers without taking a foot off the ground.