Eastern Virginia concentrates 12% of the world's data center capacity, essential for artificial intelligence. Tax exemptions attracted tech giants, but now local residents oppose the environmental impact of these facilities, which consume enormous amounts of energy and water, generating carbon emissions and water stress.
The energy cost of training an AI ⚡
Each data center requires between 30 and 50 megawatts of continuous power, comparable to the consumption of a small city. Training models like GPT-4 requires weeks of full-load computing, dissipating heat that demands water-based cooling systems, at a rate of 4 to 8 liters per kilowatt-hour. Without efficiency improvements, the electricity demand of these centers could double by 2030.
Neighbors ask for less data and more water for the lawn 💧
Virginia residents no longer know whether to complain about traffic or that the tap spits out hot air. While companies promise jobs, residents see their electricity bills rise and rivers recede. Some joke that AI will end up asking the city council for a glass of water. Ironies of progress: the cloud consumes more resources than solid ground.