The process of training a single AI model, such as LLM, can consume thousands of megawatt-hours of electricity and emit hundreds of tons of carbon. Training AI models can also result in staggering amounts of freshwater evaporating into the atmosphere to keep data centers cool, exacerbating the strain on already limited freshwater resources. These environmental impacts are expected to grow significantly, and disparities remain in the impacts on different regions and communities. The ability to flexibly deploy and manage AI computing across a network of geographically distributed data centers offers a tremendous opportunity to address the environmental inequities of AI by prioritizing disadvantaged regions and fairly distributing the overall negative environmental impacts.
Artificial intelligence is being adopted at a rapid pace across all parts of society, offering the potential to address shared global challenges such as climate change and drought mitigation. But underlying the excitement around AI’s transformative potential are increasingly large, energy-hungry deep neural networks. And the growing demand for these complex models is raising concerns about AI’s environmental impact.