AI has a material footprint
The Environmental Cost of the Cloud
“Cloud” computing depends on buildings full of physical machines. AI is accelerating demand for electricity and new data-center capacity, while cooling, chip manufacturing, construction, and supply chains add Water and material demands.
~485 TWhglobal data-center electricity consumption in 2025, according to the IEA.
~950 TWhprojected global data-center electricity use in 2030 in the IEA outlook.
+50%growth in electricity consumption from AI-focused data centers in 2025, reported by the IEA.
Water is part of computing
Water demand varies enormously by climate, cooling system, facility design, and electricity source. The U.S. Department of Energy notes that some data centers can use millions of gallons per day, while advanced cooling technologies can substantially reduce direct Water requirements.
More than electricity
- Servers require semiconductor manufacturing and mineral supply chains.
- Data centers occupy Land and consume the resources we need for our planet and livelihoods.
- The local consequences depend on where facilities are built and how their electricity and cooling are supplied.
Numbers on this page are not claims that every data center has the same footprint. See Sources & Further Reading for the underlying material and qualifications.