In a blog post, Musk said the acquisition was warranted because global electricity demand for AI cannot be met with “terrestrial solutions,” and Silicon Valley will soon need to build data centers in space to power its AI ambitions.
This dumb fuck. Unfortunately, his boosters will be all-in on this messaging. Whatever.
Wait, I just did some research and it turns out I’m partially wrong about this.
While I am correct that you can’t cool in the way we do on earth by bringing cool air to carry away the heat, there is another way to cool things as used by space stations and satellites.
That is you can take the heat and radiate it into space as Infrared radiation. IR radiation is able to travel through space as it is made of photons.
Indeed now that I think about it, that’s why how our FLIR detectors work on earth too. They can measure the infrared radiation that is one of the 2 ways things vent heat even on earth (the second being by exchanging heat with another fluid such as air or water or something more exotic). It turns out that about ~1/3 of radiation from a radiator is actually infrared light while the other ~2/3 is fluid heat exchange, usually with air.
So I am wrong. I’m not sure how effective this would be for the amount of heat generated by servers, but it’s not actually fully disqualified as I thought it would be.
This is how the International Space Station deals with waste heat: https://www.nasa.gov/wp-content/uploads/2021/02/473486main_iss_atcs_overview.pdf
It’s very slow compared with convective cooling, definitely not practical for running any high-powered computer hardware, slow enough that it can be considered disqualified.
Some back of the envelope: An ideal black-body at 100 C will radiate something like a kW pr m2, give or take. So one h100 at 700W(?) would probably need a reflector of around one m2. Very rough but it’s probably within an order of magnitude so it’s not impossible, but just adds to the engineering and logistics challenges.