Google's proposed constellation of free-fall satellites (“without thrust”); arrow pointing to Earth.
However, there is a physics problem. The resulting power decreases as the square of the distance, so Google notes that the satellites will have to maintain proximity at a distance of a kilometer or less. This would require a tighter formation than any grouping currently operating, but it should be workable. Google has developed analytical models suggesting that satellites positioned a few hundred meters apart would require only “modest position-keeping maneuvers.”
Hardware designed for space is expensive and often less efficient than ground-based systems, since the former must be protected from extreme temperatures and radiation. Google's approach to the Suncatcher project is to reuse components used on Earth that might not be very reliable if placed on a satellite. However, innovations such as Powered by Snapdragon The Mars Ingenuity helicopter showed that off-the-shelf hardware could survive in space longer than we thought.
Google claims that Suncatcher only works if the TPUs can last at least five years, which is 750 rads. The company is testing this by firing its latest Cloud TPU v6e (Trillium) with a beam of 67 MeV protons. Google says that while memory is most vulnerable to damage, experiments have shown that TPUs can withstand about three times as much radiation (nearly 2 krad) before data corruption is detected.
Google hopes to launch a couple of prototype TPU satellites by early 2027. The company expects the cost of launching these first AI orbiters to be quite high. However, Google is targeting the mid-2030s, when startup costs are projected to drop to just $200 per kilogram. At this level, space-based data centers can become as cost-effective as terrestrial versions.
The fact is that land-based data centers filthynoisy and greedy for energy and water. This has led many communities to oppose plans to build them close to where people live and work. Launching them into space could solve everyone's problems (if you're not an astronomer).
