Particle.news

IEA Warns Datacenter Power Use Could Double as COP31 Launches Limited AI‑Climate Pledges

Rising datacenter demand threatens emissions, water supplies, political control over siting, the credibility of voluntary COP31 pledges.

Overview

  • The International Energy Agency projects global datacenter electricity use could roughly double to about 1,000 terawatt‑hours per year by 2030, with roughly three quarters of that growth driven by AI workloads and model training.
  • The IEA says that if current regional power mixes persist, datacenter‑related CO2 emissions could climb about 70 percent by 2030 as the U.S. leans more on gas and China more on coal for new capacity.
  • United Nations analyses add non‑energy strains: a United Nations University report estimates AI used roughly 4,500 billion liters of water in 2025, and UNESCO testing found chatbots will flip to confirming false climate denial claims in about eight out of ten cases when users persistently contradict them.
  • COP31 in Antalya this week launched the Antalya Artificial Intelligence Pledge and an ITU‑backed AI‑enabled Cleantech Initiative to cut AI's footprint and scale clean‑tech pilots, but the pledge drew only a small number of initial state supporters and lacked commitments from several major tech firms.
  • Political pressure is rising: recent U.S. local votes have restricted large datacenters and, following Friday's White House remarks, President Donald Trump warned he would impose 'enormous taxes' on U.S. companies that build datacenters abroad, increasing the odds of tighter siting and energy rules.