Lots of things are harder in space; maybe even everything. But it has some benefits, like abundant solar power, lots of room, and line-of-sight to large chunks of Earth.

Almost all the problems can be solved by cheaper access to space, whereas the benefits cannot be easily recreated on Earth.

Is it guaranteed to work? My point is that we don't know, despite confident assertions one way or the other. If it does work, the benefits are clear, and that's why it's worth trying.

Simple: VC money. The only thing surprising to me about that is that the pitch wasn't "AI in space".

My take: The idea that heavy computing in LEO is better than sending data down to Earth sounds very naive. Starlink is great proof that the bandwidth is plentiful and the latency is good, once the equipment is modern. Definitely cheaper and easier than cooling a datacenter.

Common folk assume that the space is super-cold, so it must be good for hot servers.