Towards a future space-based, highly scalable AI infrastructure system design
It addresses the energy and scalability challenges for future AI systems by exploring a novel space-based approach, though it is incremental in applying existing technologies to a new context.
This paper tackles the problem of growing AI compute and energy demands by proposing a scalable space-based infrastructure using solar-powered satellites with TPU accelerators, achieving radiation tolerance for a 5-year mission and projecting launch costs below $200/kg by the mid-2030s.
If AI is a foundational general-purpose technology, we should anticipate that demand for AI compute -- and energy -- will continue to grow. The Sun is by far the largest energy source in our solar system, and thus it warrants consideration how future AI infrastructure could most efficiently tap into that power. This work explores a scalable compute system for machine learning in space, using fleets of satellites equipped with solar arrays, inter-satellite links using free-space optics, and Google tensor processing unit (TPU) accelerator chips. To facilitate high-bandwidth, low-latency inter-satellite communication, the satellites would be flown in close proximity. We illustrate the basic approach to formation flight via a 81-satellite cluster of 1 km radius, and describe an approach for using high-precision ML-based models to control large-scale constellations. Trillium TPUs are radiation tested. They survive a total ionizing dose equivalent to a 5 year mission life without permanent failures, and are characterized for bit-flip errors. Launch costs are a critical part of overall system cost; a learning curve analysis suggests launch to low-Earth orbit (LEO) may reach $\lesssim$\$200/kg by the mid-2030s.