The process of training a single AI model, such as an LLM, can consume thousands of megawatt hours of electricity and emit hundreds of tons of carbon. Training an AI model can also cause a surprising amount of fresh water to evaporate into the atmosphere for data center heat rejection, potentially putting further strain on our already limited fresh water resources. These environmental impacts are expected to intensify dramatically, and there continues to be growing disparity in how different regions and communities are affected. The ability to flexibly deploy and manage AI computing across a network of geographically distributed data centers offers substantial opportunities to address AI’s environmental inequities by prioritizing disadvantaged regions and equitably distributing the overall negative environmental impact.
The adoption of artificial intelligence has rapidly accelerated across all sectors of society, offering the potential to address common global challenges such as climate change and drought mitigation. But behind the excitement about AI’s transformative potential lie increasingly large and power-hungry deep neural networks. And the increasing demands of these complex models are raising concerns about AI’s environmental impact.
