Google’s latest moonshot to put machine learning in space

# TL;DR Google’s “Project Sunflower” proposes leveraging orbital data centers and abundant space-based solar power to efficiently scale AI infrastructure. ## Initial Reactions and Feasibility Upon first hearing the idea, the common reaction was: “That’s crazy. There must be reasons why it won’t work.” The team initially set out to identify those reasons, but ultimately proved the plan was feasible. ## Core Concept: Project Sunflower The core idea behind Project Sunflower is simple: How can we efficiently scale AI infrastructure using orbital data centers powered by the abundant energy from the sun? This initiative aims to use space-based solar energy to provide sustainable power for AI computing. ## Solar Efficiency Advantage One might ask: “Why not just use solar power on Earth?” However, data shows that solar panels placed in the right orbit can generate five to eight times more power than equivalent panels on Earth. This significantly boosts energy harvesting efficiency, providing stronger energy support for large-scale AI computing. ## Orbital Choice: Terminator Orbit During early planning, the team evaluated various orbits, including geostationary orbit. However, the terminator orbit was identified as the best fit for the goal because solar panels in this orbit can achieve nearly 24/7 uninterrupted sunlight, ensuring continuous energy supply. ## Satellite Clusters and TPU Chips At the smallest scale, a single satellite would carry dozens of TPU (Tensor Processing Units) chips for local AI processing. To tackle more ambitious problems, many such satellites need to connect into clusters and fly together, forming a distributed computing network. These satellite clusters are massive and must maintain tight formation. ## Formation Flying and Control Systems Flying satellite clusters involves intricate techniques in “Guidance, Navigation, and Control Systems.” Satellites must not only know their absolute positions but also their relative positions within the constellation, ensuring all satellites move in synchronized coordination. From the perspective of a central satellite in the cluster, this resembles a Viennese Waltz, with satellites orbiting the cluster center along specific trajectories for efficient collaboration. ## Google’s Legacy of Moonshot Projects Google has a long history of pursuing “moonshot” innovations. The most exciting aspect of Project Sunflower is that it pioneers a new way to support AI in creating value for humanity. By deploying computing power into space, Google aims to overcome Earth’s resource limitations and drive sustainable development in AI technology. ## Conclusion: The Ultimate Potential of Solar Energy The sun is the ultimate energy source in our solar system. When exploring how to scale solar power and utilize it in the simplest and most efficient way, building solar energy systems in space in the right orbit becomes the inevitable answer. Project Sunflower not only showcases the potential of technological innovation but also points to a new direction for the future development of AI infrastructure. Source: Google’s latest moonshot to put machine learning in space (https://www.youtube.com/watch?v=o1JK79jszqo)