AI Seeks Power Beyond Earth as Data Centers Hit Limits.

Tech firms are exploring space-based data centers because AI training requires large amounts of power and land-based facilities are reaching capacity. The article reports early tests and plans, including Alphabet's Project Suncatcher and Starcloud's in-orbit AI run.

Companies are exploring space-based computing as AI training increasingly requires large amounts of power, the article reports. On-Earth data centers are running into physical and power limits. Early experiments and proposals aim to move some AI workloads into orbit. Key developments: - The article mentions Alphabet announced Project Suncatcher, planning two test satellites for early 2027 to run AI chips using solar power. - The article mentions Nvidia-backed startup Starcloud trained an AI model in orbit in December and plans a larger GPU-cluster satellite by late 2026 with services targeted for early 2027. - The article mentions SpaceX is preparing for a possible 2026 initial public offering and said some proceeds would be directed toward space-based AI infrastructure. - The article notes remaining technical and safety challenges reported by experts, including current launch costs, chip cooling and radiation concerns, and risks from orbital debris. Summary: These developments show major firms are testing whether orbit can provide continuous power and additional scale for AI while facing cost, technical and safety barriers. Near-term milestones cited in the article include Project Suncatcher's test satellites in early 2027 and Starcloud's planned larger satellite by late 2026.