Google’s Project Suncatcher to put ML infrastructure in space (blog.google)

Google’s Project Suncatcher is set to launch a prototype satellite to test the performance of its Tensor Processing Units (TPUs) in space. This initiative, announced last year, aims to explore the feasibility of hosting scalable machine learning infrastructure in low Earth orbit, where satellites can harness up to eight times more solar power than on Earth. The project's potential significance lies in the possibility of connecting multiple satellite constellations to manage larger AI workloads, paving the way for advancements in various fields such as healthcare and scientific discovery. As part of its initial mission with SpaceX's Transporter-18, the team will investigate how TPUs withstand the extreme conditions of space, including intense vibrations during launch and radiation exposure in orbit. Initial tests show promising resilience to radiation and physical stress, but cooling remains a challenge due to the vacuum of space, where conventional airflow is absent. Future designs will involve satellites equipped with dozens of TPUs and laser interconnectivity to enable high-bandwidth communication across moving clusters. The outcome of this mission could lead to significant breakthroughs in the application of AI technologies across various industries.