Not every AI-era power saving comes from a new power plant -- Berkeley Lab reported in December 2025 that a two-year cooling retrofit at its NERSC supercomputing facility cut non-IT power draw, the electricity a data centre spends on cooling rather than compute, by 42%, saving more than 2 million kilowatt-hours a year
Berkeley Lab reported in December 2025 that a two-year cooling retrofit at its NERSC supercomputing facility reduced non-IT power consumption by 42%. This cut the electricity a data centre spends on cooling rather than compute by more than 2 million kilowatt-hours a year. The project also saved a half million gallons of water annually, reducing utility bills by approximately $200,000 per year.
Verified 2026-09-27
Track with InsiderWhy it matters
Better cooling can cut a data centre's power use, a cheaper fix than building new power plants.
More on Energy
- American Electric Power, one of the largest US utilities, which counts hyperscalers and manufacturers among its new large customers, signed fresh load commitments in the third quarter of 2026 and now expects its contracted extra load to reach 84 GW by 2031 AEP signed 15 GW of new load commitments in a single quarter
- AI data centres worldwide reached a power capacity comparable to New York State at peak demand in 2025, which Stanford's AI Index puts at 29.6 gigawatts
- OpenAI does not disclose GPT-4o's resource footprint, so Stanford's 2026 AI Index Report turned to independent researchers' estimates of the water needed to cool the servers behind it — as much as 1.58 million kilolitres a year — more than 1.2 million people drink annually
The AI boom,with receipts.
6 sourced facts every weekday, plus the Weekly Brief, AI company deep dives and so much more. Free.
Free · No card · Unsubscribe anytime · See today's issue →
Join now and you start with 100 Receipts on The Receipts Ledger — your first step toward a free year of Insider or Pro. How it works →