Particle.news

AI Data-Center Build-Out Drives Sharp Rise in Google and Amazon Emissions

Policy choices on how cloud operators count renewable power are now under debate because corporate reports and new research show AI infrastructure is increasing short-term carbon, energy and water pressures.

Overview

  • Corporate sustainability reports released early July show Google’s greenhouse-gas emissions rose about 18% in 2025 and Amazon’s rose roughly 16%, with both companies blaming rapid AI data-center construction and higher electricity use.
  • Google reported that building new data centers produced about 2.3 million tonnes of CO2e in 2025 and that supply-chain emissions climbed about 25%, while its water use for cooling jumped roughly 34% year-on-year.
  • Allianz Research estimates global data-center emissions at about 286 million tonnes of CO2 in 2025, a figure 57% higher than prior IEA-based counts, and warns emissions could reach about 643 million tonnes by 2030 without faster power-sector decarbonization.
  • A draft EU proposal seen by the Financial Times would relax strict time-and-place rules for renewable certificates and allow nuclear-backed certificates after lobbying by hyperscalers, a shift critics say would weaken emissions-accounting and risk greater short-term reliance on gas.
  • Where centers are sited and which grid supplies them matter strongly: the same compute task can produce as much as 24 times more emissions in a high-carbon grid, so decisions now will shape local water demand, power prices and whether new capacity is low-carbon or fossil-fuel based.