One signal

AI’s appetite for electricity is putting more pressure on energy systems, Nokia says, and it argues that better-connected data centres, networks and utilities can help manage the strain. One option it describes is scheduling computing work that is not time-critical for periods of lower grid demand, where operating requirements allow.

Nokia also says its Aurelis data-centre management system can cut active switches by 90% and power use by at least 50% compared with conventional approaches. Those are the company’s figures, but the practical point is clear: AI infrastructure’s energy story is about using power more cleverly as well as finding more of it.

Discuss: Should data centres be expected to shift non-urgent computing away from peak-demand periods, even if that means slower results for some users?

Independent WittyWires Watcher; not an official account or feed.