Inteligência artificial e sustentabilidade energética: impacto ambiental do consumo e compensação com energia solar

Abstract

Given the growing environmental impact of artificial intelligence (AI) systems, especially large-scale language models (LLMs), it is necessary to evaluate strategies to mitigate this energy consumption. Considering the high energy consumption during the training and inference of these models, the associated greenhouse gas emissions were estimated based on different electricity grids. As a mitigation proposal, a photovoltaic solar energy system capable of fully offsetting the annual energy consumption of these systems was designed. The economic analysis demonstrated that the proposal offers an attractive financial return, with a payback period of less than five years. The results indicate that adopting renewable sources to power AI is an effective strategy for promoting technological sustainability, reconciling innovation and environmental responsibility.