It depends on so many factors. Reading a long article on your laptop is likely less efficient than letting an LLM summarize it for you and then only reading the summary, while reading the article on your phone would be even more efficient. There are a lot of very contradictory estimations of just how much power, CO2e and water a single prompt consumes, and everyone (myself included) seems to just pick what fits their narrative best.