41
Sundar Pichai says Google will start building data centers in space, powered by the sun, in 2027
(www.businessinsider.com)
While I don't doubt that we will, at some point, have something like data centers in space, it kinda seems like a bad idea right now. Doing some searching, it looks like the cost to send something to orbit, using SpaceX's Falcon Heavy is something around $1,500/kg. 1, 2 For a server which weighs 2-3 kg, that's adding a significant cost on top of the expensive hardware costs already involved. Though, on the plus side, without the environmental impact and lawsuits from local opposition, this cost could balance out.
Then they need to deal with cooling. Keeping data centers cool is already a challenge. One of the main reasons communities have been lining up against data centers is their water usage, which is used for cooling. In space, you can't just tap into the nearest water supply. Radiative cooling sucks, sure you could just build a bigger radiator, but that's more mass you need to send to orbit, more complexity and something else you need to worry about micro-meteors slamming into. The International Space Station already uses a large, complex system for cooling and it has nothing like the internal heating of hundreds of GPUs churning out furry porn.
Lastly, maintenance is going to be a bitch. Granted, Microsoft has show that it is possible to run a lights-out data center effectively by dropping it in the ocean. Though, the fact that we don't see more of that tells me that the economics of it likely don't pencil out well compared to just paving over more farmland and ignoring the poors whining about things like fresh water.
This really seems like one of those ideas where someone needs to tell Mr. Pichai to put the bong down for a bit.
