Elon Musk: Why a 1 Terawatt AI is impossible on Earth?? "My estimate is that the cost-effectiveness of AI in space will be overwhelmingly better than AI on the ground. So, long before you exhaust potential energy sources on Earth, meaning perhaps in the four or five-year timeframe, the lowest cost way to do AI compute will be with solar-powered AI satellites. I'd say not more than five years from now Just look at the supercomputers we're building together. Let's say each rack is two tons; out of that two tons, 1.95 of it is probably for cooling. Just imagine how tiny that little supercomputer is Electricity generation is already becoming a challenge. If you start doing any kind of scaling for both electricity generation and cooling, you realize space is incredibly compelling Let's say you wanted to do 200 or 300 gigawatts per year of AI compute. It's very difficult to do that on Earth. The US average electricity usage, last time I checked, was around 460 gigawatts per year. So, if you're doing 300 gigawatts a year, that would be like two-thirds of US electricity production per year. There's no way you're building power plants at that level And then if you take it up to a Terawatt per year, impossible. You have to do that in space In space, you've got continuous solar. You don't need batteries because it's always sunny. The solar panels actually become cheaper because you don't need glass or framing, and the cooling is just radiative"