WHY COMPUTE MIGHT GET 10X+ MORE EXPENSIVE IN COMING YEARS

Summary from the TLDR newsletter

Labs do not want to spend greater and greater shares of compute on inference. The point of inference revenue is to convince investors to spend more money to buy more compute to train bigger models. Spending too much on compute on inference signals that progress has stalled because it's not worth investing more in training. Labs believe that their models will continue to get better, so this only means that compute will get more expensive.