‘Greetings, earthlings’: Nvidia-backed Starcloud trains first AI model in space as orbital data center race heats up (www.cnbc.com)

Starcloud, a startup backed by Nvidia, has made history by training an AI model in space for the first time using its Starcloud-1 satellite, equipped with a groundbreaking Nvidia H100 GPU. This satellite is running Google's open large language model, Gemma, demonstrating the potential for orbital data centers to handle complex AI tasks traditionally confined to Earth. With power consumption projected to explode in terrestrial data centers, Starcloud's initiative could alleviate pressure on global infrastructure by utilizing the unlimited solar energy available in space. CEO Philip Johnston emphasized the monetary and environmental advantages of space-based AI operations, asserting that these data centers will have significantly lower energy costs and can perform tasks on par with those on Earth. The model's operation also showcases its flexibility, as it engages in real-time data analysis, which could be crucial for applications such as wildfire detection and maritime search and rescue efforts. Despite the challenges of space operations—like radiation exposure and maintenance—the push toward orbital computing is gaining momentum, indicating a transformative shift in how AI workloads could be managed in the future.