Nvidia-backed Starcloud has achieved a significant milestone by training an AI model in space using its Starcloud-1 satellite, equipped with a cutting-edge Nvidia H100 GPU. This satellite is the first to run a large language model (LLM) in orbit, successfully executing NanoGPT, which was trained on Shakespeare’s works. In a historic first, Starcloud also runs Google’s Gemma model, demonstrating that advanced AI applications can operate beyond Earth. CEO Philip Johnston claims that these orbital data centers could cut energy costs by tenfold compared to terrestrial counterparts, directly addressing the growing energy demands of data centers on our planet. Starcloud envisions building a 5-gigawatt orbital data center that could harness solar energy continuously, avoiding the limitations of Earth’s weather and day-night cycles. Johnston highlighted the potential for real-time intelligence applications, such as detecting wildfires or monitoring maritime activities. Future satellites will incorporate more Nvidia chips and a cloud platform from Crusoe, enhancing AI capabilities from space. However, analysts point out potential challenges, including radiation exposure, maintenance difficulties, and regulatory hurdles, which could complicate the operation of these data centers. Despite these risks, interest in orbital data centers is rising among major tech companies. Google is pursuing its Project Suncatcher to deploy solar-powered satellites, while other firms plan lunar and orbital data center projects. Starcloud’s collaboration with SpaceX for satellite launches further emphasizes the growing momentum in this sector.