Picture groups of satellites in a low-Earth orbit that is synchronized with the sun, the very source of the energy that powers them. Then imagine each satellite outfitted with AI chips doing the business of machine-learning high above our planet. The satellites are between 100 and 200 metres apart within a cluster that is no more than 1 kilometer wide but they have optical connections that enable them to serve as a single data center.

This is Project Suncatcher, a vision announced by Google earlier this month. It exists only in theory - at least for now. Google has teamed up with Planet, an Earth-imaging company, in the hopes that the partnership can launch a so-called "learning mission" into space in 2027 using two prototype satellites to test the technology and the hardware.

Meanwhile, Google is working out the costs of Project Suncatcher when it ultimately deploys: The launch alone is considered to have a prohibitive price tag. There are also issues of aerospace engineering that need to be resolved to keep the satellite cluster in good repair.

With AI's established reputation as a voracious consumer of energy, the company says it was only logical to look beyond the horizon - way beyond the horizon - to get closer to a source of power. Just what is the prospect for the success of Project Suncatcher? Right now, you won't find that answer, even if you google it.