Google and SpaceX are in talks about putting data centers in orbit, according to a new report. The pitch is straightforward: space could be the next frontier for AI compute infrastructure.
The economics don't work yet. Building and operating data centers in space costs significantly more than traditional ground-based facilities. But the companies are apparently betting that could change as launch costs drop and AI's power demands keep climbing.
For AI companies, the appeal is obvious. Space offers unlimited cooling (just radiate heat into the void) and access to solar power 24/7 without weather interruptions. No need to worry about local power grids or water supplies for cooling towers.
The timing makes sense given where AI infrastructure is headed. Training runs are getting more expensive, and finding enough power and cooling capacity on Earth is becoming a real constraint. Major AI labs are already scouting locations based on energy availability.
Still, this is very early stage talk. Launching hardware into orbit, maintaining it remotely, and dealing with latency for Earth-based users are all unsolved problems at scale. But if you're planning AI infrastructure for the 2030s, apparently you need to think beyond just building another data center in Iowa.
