Google's ambitious Project Suncatcher aims to revolutionize AI by placing powerful Tensor Processing Units (TPUs) in space. This groundbreaking initiative, detailed in their research blog, envisions a network of interconnected satellites harnessing the Sun's energy to scale machine learning capabilities. With solar panels up to 8 times more productive than on Earth, these satellites will operate in a 'dawn-dusk sun-synchronous low Earth orbit', reducing the need for traditional power generation methods. The key to this concept lies in free-space optical links, enabling high-bandwidth, low-latency connections between satellites, akin to data centers on Earth. However, this requires satellites flying in very close formation, just hundreds of meters apart, demanding precise station-keeping maneuvers. Google's initial radiation testing on TPUs, known as Trillium, v6e, revealed promising results, with no hard failures observed up to 15 krad(Si) of radiation exposure. The company predicts launch costs will decrease to under $200/kg by the mid-2030s, making space-based data centers economically viable. While Google acknowledges engineering challenges like thermal management and reliable ground communications, they are partnering with Planet to launch prototype satellites by early 2027, marking a significant step towards a future where space-based AI infrastructure becomes a reality.
