Benedetta Brevini
Visiting Professor at New York University, Institute of Public Knowledge and Associate Professor of Political Economy of Communication at the University of Sydney
Benedetta Brevini is Visiting Professor at New York University, Institute of Public Knowledge and Associate Professor of Political Economy of Communication at the University of Sydney. Before joining the academy, she worked as journalist in Milan, New York and London for CNBC, RAI and the Guardian. She is the author of several books including Is AI good for the Planet (2022), Newscorp: Empire of Influence (2024), Amazon: Understanding a Global Communication Giant (2020), Public Service Broadcasting online (2013) and the editor of Beyond Wikileaks (2013), Carbon Capitalism and Communication: Confronting Climate Crisis (2017), Climate Change and the Media (2018). She is currently working on a new volume for Polity entitled “Communication systems, Technology and the climate emergency”.
Much more than data centres’ energy demands: exploring an Eco-Political Economy of AI to Understand its environmental toll
In the lead-up to the much-anticipated Climate Summit (COP 29) in Baku, Azerbaijan, stories about the unsustainability of the data centers that power AI are making headlines. It’s about time the environmental impact of AI came under scrutiny. In the past six months, the conversation has shifted dramatically, making it almost impossible for the AI industry to avoid discussing AI’s massive environmental costs. So, what has changed?
Certainly, the staggering rise in energy and water consumption driven by generative AI, that has forced major Digital Lords to be more open about this dire issue, which can no longer be ignored. Microsoft's most recent environmental report following the launch of OpenAI's generative AI services reveals a substantial 34% increase in global water consumption from 2021 to 2022, reaching nearly 1.7 billion gallons. Goldman Sachs predicts that data centers will use 8% of US energy by 2030, compared with 3% in 2022.Additionally, the Intergovernmental Panel on Climate Change (IPCC), made up of the world’s leading climate scientists, released its last report in March 2023. Defined by many the ‘Final Warning’, this is the last assessment while the world still has a chance to limit global temperature rises to 1.5C above pre-industrial levels, the threshold beyond which our damage to the climate will become irreversible. To keep within the 1.5°C limit, emissions need to be reduced by at least 43% by 2030 compared to 2019 levels, at least 60% by 2035. These targets seem to be moving progressively far away.
While this change in public discourse might be seen as good news, especially for those of us who have been sounding the alarm about the environmental costs of AI technologies for years, I argue that it is crucial to move beyond reductive frameworks that focus solely on carbon and water footprints. Instead, we must understand holistically the complexities of the environmental harms of AI. In various interventions, I have called for the development of what I termed an "Eco-Political Economy of AI" (Brevini, 2021; 2022; 2024) to address the complex factors involved in assessing AI's environmental impacts. This approach involves examining three critical segments of the extractive global production and supply chain of AI to account for its environmental costs: a) mining and resource extraction; b) consumption, energy use, and carbon footprints; and c) digital waste. In this talk, I will delve deeper into these ideas, presenting a nuanced elaboration of this framework. By integrating theories from media and communication, geography, computing, and engineering with indigenous concepts and environmental justice paradigms, the Eco-Political Economy of AI offers the tools necessary to place the climate crisis at the forefront of technology and AI development.