AI data centres have massive electricity demands, which are estimated to quadruple over the next decade.Credit: Mario Tama/Getty.
Ten countries including Germany and Brazil at the COP31 climate summit in Antalya, Turkey, have pledged to do more to limit the greenhouse-gas emissions produced by the rapid development of artificial intelligence — and to use the technology to cut emissions elsewhere.
In a recent development, how much energy will AI really consume? The good, the bad and the unknown.
It has been revealed that how much energy will AI really consume? The good, the bad and the unknown.
“The pledge brings together two priorities that must advance together: reducing the environmental footprint of AI and expanding its contribution to climate action,” stated Mehmet Fatih Kacır, Turkey’s minister of industry and technology at a launch event for the Antalya Artificial Intelligence Pledge on 16 November.
In a recent development, some climate advocates welcomed the pledge as a necessary, if vague, step to address the growing emissions and environmental impacts caused by running AI data centres. And researchers say there are ways that the technology could be used to cut emissions, for instance by helping to integrate renewables onto power grids or by improving logistics.
But it remains unknown whether such emissions reductions could offset AI’s consequences for the climate. “The trillion-dollar question is whether or not those savings will outweigh those impacts we’re seeing now,” indicates Eric Masanet, a mechanical engineer who studies decarbonization at the University of California, Santa Barbara.
It has been revealed that the rapid building of AI data centres — which are used to train and operate AI systems — has drawn increasing ire from environmental quarters because of its insatiable demand for electricity. A report this month from BloombergNEF, which provides research on the clean-energy transition, estimated that electricity demand from data centres will quadruple over the next decade, making up 7% of global electricity demand in 2035, and accounting for at least 2% of global emissions.
According to recent reports, most of that would come from the United States, where the building of data-centres has been biggest and fastest, followed by China. In the race to train the most capable models, many big tech firms in the United States are seeking to build gas-fired power plants to satisfy data centres’ energy demands, rather than waiting for permission to connect to the public power grid.
Can AI models reliably forecast extreme weather events?
Can AI models reliably forecast extreme weather events?
Recent findings suggest that that’s the sort of approach the pledge aims to prevent. It calls on countries to take action to power AI infrastructure with “zero- and low-emissions energy” and to improve the efficiency with which data centres use power and water for cooling. Countries that join will assess progress yearly.
The other G20 countries that signed on are South Korea and Turkey, which sponsored the pledge and mentioned it would aim to power its data centres with 60% low-carbon sources. They were initially joined by Andorra, Colombia, Luxembourg, Sierra Leone, Somalia and Uganda. More than 80 organizations endorsed the pledge, including a few large tech companies such as Nokia and Telefónica, according to a tracker from the UN’s International Telecommunications Union, one of the effort’s organizing bodies. However, none of the frontier AI firms have joined the pledge.
Given the absence of AI’s biggest emitters, some climate advocates dismissed the initiative. “The AI pledge risks giving Massive Tech a green stamp without requiring it to address the real climate impacts of its infrastructure,” says Jill McArdle, a campaigner at Beyond Fossil Fuels, an advocacy group in Berlin.
doi: https://doi.org/10.1038/d41586-026-03763-z.
Stern, N. et al. npj Climate Action 4, 56 (2025).
According to recent reports, alpine, W. et al. npj Climate Action 5, 71 (2026).
Net zero needs AI — five actions to realize its promise.
Light bulbs have energy ratings — so why can’t AI chatbots?
Data centres will use twice as much energy by 2030 — driven by AI.
Will AI accelerate or delay the race to net-zero emissions?
As El Niño builds, studies of extreme marine heatwaves must start now.
Biotech innovations are an opportunity to limit climate change.
Can green technologies deliver energy security in an unstable world?
AI’s energy paradox: will the AI boom strain energy systems or be the source of solutions?
Can green technologies deliver energy security in an unstable world?
Flexible rubber blades for tidal turbines.
AI wearables could exploit people’s most vulnerable moments.
AI’s energy paradox: will the AI boom strain energy systems or be the source of solutions?
Can green technologies deliver energy security in an unstable world?
NBIC, Peking University, seeks Assistant Professors and Research Track faculty in imaging, probes, spatial omics and digital life. Open until filled.
National Biomedical Imaging Center, College of Future Technology, Peking University.
Job Title: Deputy Editor, Nature Progress Bioengineering Location: Shanghai, Beijing, New York, Philadelphia - hybrid working model Closing date: 2...
Shanghai, Beijing, Upcoming York, Philadelphia - hybrid working model.
The SoQ invites highly qualified faculty member in quantum sciences & technology and related fields.
School of Quantum at the University of Chinese Academy of Sciences (UCAS).
Join HZAU's global faculty team to advance research with competitive benefits.
Huazhong Agricultural University (HZAU).
The School of Life Science and Technology (SLST) invites applications for numerous tenure-track and tenured positions.
