Soaring demand for AI chips fuels power usage: report
Global chip manufacturing is concentrated in East Asia, including Taiwan, South Korea and Japan, whose power systems are still heavily dependent on burning fossil fuels that produce climate-warming emissions.
Exploding demand for the highly advanced chips used in energy-hungry AI data centres around the world has raised fears about the technology's environmental impact.
AI chipmaking used nearly 984 gigawatt hours (GWh) last year, up more than 350 percent from 2023, Greenpeace said.
Emissions from that electricity generation rose more than four-fold to 453,600 metric tons of carbon dioxide equivalent in 2024, from 99,200 in the previous year.
"Chip manufacturing is extremely energy intensive. A large foundry can consume up to 100 megawatt-hours of electricity per hour," Greenpeace said.
The group warned that global electricity demand for AI chipmaking was expected to reach 37,238 GWh by 2030, adding that was "more than the total current electricity consumption of Ireland".
Tech companies were spending billions of dollars in AI research and infrastructure, but "the adoption of strategies to achieve 100 percent renewable energy across supply chains remains limited", Greenpeace said.
Increasing demand for chips and expanding manufacturing capacity would "inevitably contribute to a growing carbon footprint", it warned.
"Tech companies such as Nvidia, Microsoft, Meta, and Google must support their suppliers to increase renewable energy procurement and should target 100 percent renewable energy across their supply chains by 2030," the report said.
While chipmakers, including Taiwan Semiconductor Manufacturing Co (TSMC), have been transitioning to renewable energy, the pace was "slow", Greenpeace said.
TSMC told Bloomberg News that the company was committed to low-carbon manufacturing and that its internal tally showed emissions per unit declined in 2024.
Related Links
Computer Chip Architecture, Technology and Manufacture
Nano Technology News From SpaceMart.com
