The Environmental Impacts of Machine Learning Training Keep Rising Evidencing Rebound Effect

TL;DR

This study investigates whether advances in hardware efficiency and algorithmic optimization can curb the environmental footprint of AI model training. It employs a bottom-up MLCA framework and a newly assembled dataset on graphics-card production to quantify impacts across hardware production and training use over 2013–2025. The results show training-related environmental footprints rise exponentially despite improvements in hardware and methods, driven by rebound effects and ongoing production emissions. Carbon-optimization strategies offer only short-term relief, underscoring the need to reduce AI activities and reconsider the scale and frequency of resource-intensive training.

Abstract

Recent Machine Learning (ML) approaches have shown increased performance on benchmarks but at the cost of escalating computational demands. Hardware, algorithmic and carbon optimizations have been proposed to curb energy consumption and environmental impacts. Can these strategies lead to sustainable ML model training? Here, we estimate the environmental impacts associated with training notable AI systems over the last decade, including Large Language Models, with a focus on the life cycle of graphics cards. Our analysis reveals two critical trends: First, the impacts of graphics cards production have increased steadily over this period; Second, energy consumption and environmental impacts associated with training ML models have increased exponentially, even when considering reduction strategies such as location shifting to places with less carbon intensive electricity mixes. Optimization strategies do not mitigate the impacts induced by model training, evidencing rebound effect. We show that the impacts of hardware must be considered over the entire life cycle rather than the sole use phase in order to avoid impact shifting. Our study demonstrates that increasing efficiency alone cannot ensure sustainability in ML. Mitigating the environmental impact of AI also requires reducing AI activities and questioning the scale and frequency of resource-intensive training.

Figures (13)

• Figure 1: Hardware modeling in MLCA.
• Figure 2: Scope of environmental assessments of training in this study. The barrel represents fossil fuels, the droplet represents water, the pickaxe represents metals, and the plug represents electricity. The top row pictograms represent, from left to right, emissions in soil, water and air.
• Figure 3: Evolution of the characteristics of NVIDIA workstation graphics cards from 2013 to 2025. Dot size represents memory size and color represents GPU technological node.
• Figure 4: Evolution of the energy consumption of NVIDIA workstation graphics cards (2013-2025).
• Figure 5: Evolution of the production impacts of the graphics cards used for training systems in the Epoch AI database. Value intervals correspond to cases of ambiguous card names.