Cette étude porte sur les réserves rentables mondiales de métaux, et notamment ceux utilisés pour fabriquer nos équipements numériques et high-tech. Les pays occidentaux sont de plus en plus dépendants des hautes technologies (high-tech). Or, les « stocks » de minerais permettant de fabriquer ces technologies, dont le numérique et l’intelligence artificielle, semblent se dégrader.
As computational demands continue to rise, assessing the environmental footprint of AI requires moving beyond energy and water consumption to include the material demands of specialized hardware. This study quantifies the material footprint of AI training by linking computational workloads to physical hardware needs. The elemental composition of the Nvidia A100 SXM 40 GB GPU was analyzed using inductively coupled plasma optical emission spectroscopy, identifying 32 elements. Results show that AI hardware consists of about 90% heavy metals and only trace amounts of precious metals — copper, iron, tin, silicon, and nickel dominate by mass. Using a multi-step methodology integrating these measurements with computational throughput per GPU across varying lifespans, the study finds that training GPT-4 requires between 1,174 and 8,800 A100 GPUs depending on Model FLOPs Utilization (MFU) and hardware lifespan — corresponding to the extraction and eventual disposal of up to 7 tons of toxic elements. Combined software and hardware optimization strategies can significantly reduce material demands: increasing MFU from 20% to 60% lowers GPU requirements by 67%, extending lifespan from 1 to 3 years yields comparable savings, and implementing both measures together reduces GPU needs by up to 93%. The study highlights that incremental performance gains — such as those between GPT-3.5 and GPT-4 — come at disproportionately high material costs, and underscores the necessity of incorporating material resource considerations into discussions of AI scalability.
L’objectif de cette étude et d’apporter un éclairage scientifique par une évaluation quantifiée des impacts environnementaux du numérique, afin que chacun∙e d’entre nous, citoyen, entreprise, dirigeant politique, puisse prendre la mesure des impacts du numérique et prendre nos responsabilités pour réduire ces impacts.