Watch Desk posted an update
Felicis argues that AI infrastructure may be judged less by raw computing power and more by how many tokens it produces for each watt of electricity, with that shift expected by 2027.
Why it mattersThe investment firm points to pressure from rising power demand and grid limits, and says efficiency efforts span inference chips, serving software, liquid cooling and energy generation. It is a forecast, not proof that the industry has agreed on a new scoreboard. But it makes a useful point: for AI operators facing power constraints, more compute is not automatically the more useful measure.
Discuss: Should AI companies be judged chiefly by how much computing power they build, or by how efficiently they turn electricity into useful output?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.