LLM inference uses on the order of 1 Wh per query. That's under 10 meters of driving on an EV or running air conditioning for under 5 seconds.
replyhttps://hannahritchie.substack.com/p/ai-footprint-august-202...
One query is not going to be a useful benchmark when people are deploying AI swarms in loops to solve simple problems
replyOr a human riding a stationary bike for 36 seconds.
reply