AI Data Centers in Space? Big Tech Thinks It’s Time
Google and SpaceX are negotiating to launch orbital data centers that could revolutionize AI computing infrastructure according to Wall Street Journal reports. The partnership would unite two tech giants competing in the emerging space-based computing race. However, experts question whether the economics make sense given current launch costs of $7,000 per kilogram.
Google’s Project Suncatcher announced November 2025 aims to launch prototype satellites by 2027 equipped with Tensor Processing Units. The moonshot explores equipping solar-powered satellite constellations to scale machine learning compute in space. Furthermore, Google is talking with other launch providers including Planet Labs for satellite design and construction.
SpaceX pitches orbital facilities as the cheapest AI compute option within years as it prepares a record-breaking IPO. Elon Musk claims space-based infrastructure avoids neighborhood opposition plaguing U.S. data center construction while leveraging constant solar power. Meanwhile, Google CEO Sundar Pichai stated in November that orbital data centers will become normal within a decade.
The deal follows SpaceX acquiring xAI in February 2026 announcing plans for 1 million orbital data satellites. Anthropic recently signed an agreement accessing xAI’s Memphis supercluster with potential orbital collaboration. Consequently, Tom’s Hardware analysis shows Project Suncatcher requires launch costs around $200 per kilogram versus current $2,700-$7,000 rates.
Critics including OpenAI’s Sam Altman call orbital data centers “ridiculous for now” citing satellite construction costs and launch expenses. Engadget experts warn cosmic radiation affects GPU error-free calculations while heat dissipation in near-vacuum space remains challenging. Global data center electricity consumption is projected to exceed 1,000 terawatt-hours by end of 2026 equivalent to Japan’s annual consumption.





