The Hidden Cost of AI: Carbon Footprint and Mitigation Strategies

Abstract

The integration of Artificial Intelligence (AI) into modern economies holds transformative potential, but its environmental impact, particularly its carbon footprint, is a growing concern. This paper explores the hidden costs associated with AI, focusing on its substantial greenhouse gas (GHG) emissions. Training large-scale AI models, particularly those based on deep learning, is highly energy demanding, resulting in substantial GHG emissions. Operationally, the continued use of AI systems further exacerbates the environmental toll, especially in data centres powered by non-renewable energy sources. This paper highlights mitigation strategies, including the transition to renewable energy sources for powering AI infrastructures and the development of more energyefficient algorithms. Techniques such as model pruning, quantisation, and knowledge distillation are identified as crucial in reducing energy consumption during the training and operational phases of AI models. Additionally, the role of AI in sustainability efforts is examined, suggesting that AI could facilitate resource efficiency in industries such as agriculture, commerce, and manufacturing, thereby contributing to the global transition towards a low-carbon economy. While AI promises significant advancements across multiple sectors, it is essential to address its environmental costs through sustainable practices. Failure to do so may result in AI accelerating climate change, overshadowing its potential benefits.

Keywords: AI, carbon footprint, climate change, cost, green AI

JEL Classification: Q54, Q55, Q56, O33, F64