dug 
Responding to a post but thoughts too wordy for LI comments field…
We’re locked in a culture of grow, more, more, faster, and design (all disciplines) has obviously been drawn into that. Your concerns are real. Looking at what data-centres are doing to communities in the USA is pretty chilling. The economics and the power distribution are all heading quickly into dystopia territory, it’s pretty nasty stuff.
I wonder if another way might work.
While I constantly fall in love with the capabilities of new models, older, smaller (cheaper) models can already do things that add amazing value to human-computer interaction and user experience generally. These benefits are less ‘showy’ (things like being able to understand broken data to some extent; being able to perform simple transpositions in .csv files; being able to generate suggestions from lists based on inputs) but they offer genuine UX improvements with much less damage. Struggling to get a perfect wireframe from Figma Make doesn’t seem like a good use of that energy and chasing the showy stuff often adds latency, cost, and failure modes for marginal gain.
Coming up with a way we can help customers get closer to the positive outcomes we’re regulated to deliver them and scribbling that on a napkin and then hacking a quick proto together seems much more exciting…