Failing to present the numbers in context is problem. Say a Google query takes approximately D drops of oil after you figure everything in. Is that a lot or a little?

For it to be meaningful you have to know how time spent on GPT compares in terms of energy expenditure to the average energy expenditure of the alternative ways people would spend time. If you do the exercise, it's hard to find any plausible scenario in which GPT, or internet time in geners, is not an energy saver!

For just a single example, there are about 34KWh of energy in a gallon of gas. Say you get 20 miles/gallon. If a query takes 14 Wh. That's 2428 queries/gallon, i.e. 6.6 queries/day, 7 days a week, all year.

If you figure an average of 10 seconds to type out the queries, it would take 6.7 hours of pure typing to use one gallon of gas worth of GPT. Figure in lunch, and a meeting, and it's a work day just for typing.

Now say, to actually use GPT, you spend 10x as much time thinking, formulating the question, writing down the answer, etc. as you spend actually typing the question. (Pick whatever number you find plausible.) That means GPT would be soaking up eight work-days of your time for the amount of energy in one gallon of gas.

Now consider that Americans burn about 1.36 gallons of gas per day, on average. Eliminate GPT, and give your super-heavy user back 8.5 work days of free time to burn energy in other ways. If he or she uses just a tiny fraction of that time to burn a gallon of gas to drive to one concert, or go to the mall, and you've now exceeded the entire energy cost of GPT for the year.

This is true for almost all Internet use. The energy consumption sounds staggering until you compare it to almost anything else people would do with the free time.