Workshop on AI & Sustainability

Behind the rapid progress of artificial intelligence lies a growing concern: the environmental footprint of AI systems. From the energy demands of large-scale model training to the extraction of critical materials for hardware and water consumption, AI’s ecological impact raises complex questions about sustainability and governance. This event brings together scholars and experts to examine the multiple facets of the environmental footprint of artificial intelligence.

Programme

9:00 - 9:30: Welcome Coffee

9:30 - 9:40: Welcome Remarks
• Constance de Leusse, Executive Director, AI and Society Institute
• Hugo Mercier, Scientific Director, AI and Society Institute
• Aurélie Bugeau (Full Professor, University of Bordeaux, LaBRI) / Anne-Laure Ligozat (Full Professor, ensIIE, LISN), Scientific Directors, Observatory on AI’s Environmental Footprint

9:40 - 10:10: Marlène De Bank, Research Engineer, The Shift Project - Current and Future Carbon Emissions of AI Infrastructures and Data Centers

Abstract: This presentation examines the explosion in computing power, which leads to the expansion of the data centre industry and computing infrastructure, and to the explosion in the supply of generative AI services. The objective is to identify the key energy and climate challenges at each level : global, European, French national and regional and to formulate recommendations enabling our technological choices to be aligned with climate objectives.

10:10 - 10:55: Gauthier Roussilhe, PhD Candidate, Royal Melbourne Institute of Technology (RMIT) - How to Build and Use Computing Power Faster : an Economic and Environmental Perspective of GPU for GenAI

Abstract: Generative AI development has been partly unlocked by the efficiency gains of modern GPUs and the increase in the production capabilities of such hardware. But what do we know of the footprint of GPUs? Are they different from other computing components and if so, to which extent? The first part of this talk will give an overview of GPU manufacturing chains and how more power consuming GPUs have changed broader supply chains. However, can we keep manufacturing GPUs at a steady growth rate? Will we have a sustained demand? In a second part, the presentation will explore how the semiconductor industry economics meet with genAI economics and the challenges that arise for future GPU demand.

10:55 - 11:15: Break

11:15 - 11:45: Ana Valdivia, Research Lecturer in AI, Government & Policy, Oxford Internet Institute - Title: The supply chains of GenAI: scale, knowledge, and relationality

Abstract: In this talk we are going to explore the supply chains of GenAI, from silicon to e-waste landfills, from a transdisciplinary perspective. Drawing on STS, political ecology, and computer science, this presentation explores the evolution of global production networks of GenAI from a scale, knowledge and relational perspective, evaluating the environmental impacts and occupational health risks related to this technology and its infrastructure. By unravelling each stage of the production network, the talk unveils continuous and discontinuous frictions emerging from the ecological ruins of the production of GenAI today. In the current context of GenAI growth and global warming, it advocates for blending the boundaries between disciplines by combining both computational and ethnographic methodologies to investigate AI supply chains and its infrastructure, for eventually driving our society towards a just transformation.

11:45 - 12:25: Sasha Luccioni, Climate Lead, Hugging Face - Title: AI for Sustainability and the Sustainability of AI: Two Sides of a Different Coin

Abstract: Humanity is currently facing an unprecedented ecological crisis, and the role of AI in this crisis remains to be defined and analyzed. On the one hand, AI can be used to benefit the environment—for protecting biodiversity, improving climate prediction, and adapting to climate change. On the other hand, the development, training, and usage of AI models have their own environmental impacts. This presentation will address both sides of this debate, explain why they are wrong, and propose ways to make AI truly sustainable.

12:25 - 12:30: Concluding remarks
Théophile Lenoir, Coordinator, Observatory on the Environmental Footprint of AI, AI and Society Institute

Pour en savoir plus : https://www.ens.psl.eu/agenda/workshop-ai-sustainability/2025-12-10t080000