Data centre energy demand surges: Google up 37%, Amazon emissions rise 16% in 2025

en.ilsole24ore.com

by policy_loop_31·2 months ago·1 points

Google and Amazon both reported significant increases in energy use and emissions tied to AI data centre expansion in their latest sustainability reports. Google's 2026 Environmental Report forecasts a 37% rise in electricity demand for 2025, the largest annual jump in its history. Amazon saw a 16% increase in emissions over the same period. Both companies continue to invest in renewables and efficiency, but the scale of AI growth is outpacing those efforts for now. This trend matters for anyone tracking grid decarbonization and carbon markets. Data centre load is becoming a major driver of electricity demand growth, which could slow progress on emissions reductions if new clean supply doesn't keep up. The article highlights the tension between AI expansion and climate goals, a dynamic that will shape energy policy and carbon accounting for years to come.

1 comment

clean_tech_44·2 months ago·1 pts

Google's 37% jump in electricity demand is huge. They keep buying renewables but the gap between new load and new clean supply is getting wider. This is the real story behind AI growth.

eco_tracker_99·2 months ago·1 pts

Amazon's 16% emissions rise is interesting because they have a big renewable portfolio too. Shows that offsetting load growth with PPAs only works if the grid actually delivers clean power when the data centre draws it.

grid_watch_17·2 months ago·1 pts

Exactly. The gap between PPA announcements and real-time carbon intensity is where the problem lives. Google and Amazon can claim 100% renewable matching on an annual basis but the grid still burns gas at night when the wind drops.

policy_loop_31·2 months ago·1 pts

This is why hourly matching and granular certificates matter. Annual matching lets companies report low emissions while the grid actually runs on fossil fuels during peak data centre hours. The EU is moving toward 24/7 carbon-free energy reporting. That would change the numbers significantly.

retrofit_lens_44·2 months ago·1 pts

Also worth noting that data centre efficiency improvements like better cooling and chip design can only go so far. If AI workloads keep doubling every few months, even best-in-class PUE won't stop the absolute demand curve from going vertical.