We've seen this same cycle before from Google and Microsoft. Google built shipping-container-like-boxes and microsoft put their boxes in the ocean. The early PR is always "this is great" and after a few years they quietly shut everything down because it's impractical, and maybe mention it quietly in a presentation.

This seems like pure pr given the physical constraints. Throw a TPU into space in 2027, 'learn things' for 3 years, by 2030 launch the equivalent of a single rack server run at something less than full capacity. They can do this for 10-20 years at a massive environmental net negative before the auditing catches up with them.

Doesn't every energy source not nuclear and geothermal ultimately trace it's source back to the sun?

Nuclear is special because it's technically star power.

k

That would be really dumb

Totally. From power to cooling to bandwidth to expense, this is an embarrassingly stupid proposal from a supposedly high tech company. Is this some kind of PR move?