Data centers around world already guzzle electricity,water like some of biggest nations, United Nations University report revealed June 3 in Washington. Their environmental footprint—energy use,water consumption,pollution—is set to double by 2030 as AI booms.
Last year,data centers consumed an estimated 448 trillion watt-hours (TWh) of electricity. This topped Brazil's total of 410 TWh in 2023. By 2030,demand could hit 800 TWh to 1,000 TWh,rivaling countries like Japan (980 TWh in 2022) or Germany (480 TWh in 2022).
Water usage also raises alarms . In 2022,data centers used 1.7 billion cubic meters (m³) of water,more than Saudi Arabia's 1.69 billion m³ same year. Projections say it could soar to 2.8 billion m³ to 3.8 billion m³ by 2030,akin to Canada's 3.4 billion m³ in 2019 or Australia's 2.9 billion m³ in 2019.
Carbon emissions from data centers hit between 240 million and 340 million metric tons of CO2 equivalent in 2022. More than Brazil's 215 million tons that year. Report forecasts emissions might rise to 320 million to 620 million metric tons by 2030.
AI models major driver here. Training large models,like OpenAI's GPT-3,used about 700,000 liters of fresh water. Adds to environmental load.
Many data centers lean on fossil fuels, worsening carbon footprint. Cooling systems,water-heavy,are problematic in dry regions where centers sometimes sit. These issues pile on environmental stress.
Report suggests fixes. Moving data centers to cooler spots or using renewables could help. Better cooling efficiency,liquid cooling,recycling waste heat offer solutions .
Smarter AI algorithms key. Policymakers could use carbon taxes,water pricing,energy standards to drive change. More transparency from tech firms needed…
