World news and discussion on TalkFuse.
US hyperscale data centres used 1.8% of electricity in a year
A Harvard-led study estimates that 403 US hyperscale data centres used 1.8 per cent of the country’s electricity between May 2024 and April 2025. Their demand shows how quickly large-scale computing has grown, while raising questions about the electricity and emissions associated with cloud and AI services.
The researchers estimated annual use of 68 to 99 terawatt-hours across four operating scenarios, comparable at the upper end to the electricity consumed by a state such as Indiana. Estimated emissions ranged from 21 million to 31 million metric tonnes of carbon dioxide, with a central estimate of 25.7 million, against 10.5 million tonnes in 2018. About 54 per cent of the electricity attributed to the sites came from fossil fuels; the study notes its estimates are limited by incomplete site data and do not distinguish AI workloads at individual campuses.
In short
- 403 hyperscale data centres used an estimated 1.8% of US electricity.
- Their estimated emissions reached 25.7 million tonnes of CO₂.
- Researchers say the figures may have risen as facilities expand.
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Hyperscale data centres support cloud computing, AI research and digital services that can improve productivity and enable new products. Their electricity use may be justified if the benefits are substantial, and operators can reduce the impact by improving efficiency, adding low-carbon power and locating facilities where the grid can support them.
The case against
The scale of electricity demand and emissions raises a reasonable concern that rapid data-centre growth could strain power systems and slow emissions reductions, especially while about half of the estimated electricity comes from fossil fuels. Because the study’s estimates are uncertain and do not isolate AI workloads, policymakers and communities may want clearer reporting and stronger safeguards before further expansion.
Explained simply
Hyperscale data centres are very large facilities that house computer servers and equipment. They support cloud services, which let people and organisations store data and use software over the internet, as well as computing tasks such as artificial intelligence.
The key players include the companies that build and run these facilities, the electricity suppliers that power them and researchers studying their effects. The Harvard-led study estimates use across 403 US sites, but says incomplete information about sites limits its figures.
The issue matters because electricity demand from data centres is growing alongside the use of cloud and AI services. How that electricity is generated affects the emissions linked to those services, while estimates cannot yet show how much power individual sites use for AI.
Americas Climate Research Science World
Read the full article at the source →
Originally published by The Register as “America’s hyperscale bit barns draw as much power as entire states”.