Improving the Environmental Footprint of AI Hyperscalers by Andrea Marinoni

Abstract:
The rapid growth of artificial intelligence, driven by large language models and other foundation models, demands highly reliable power infrastructures. Yet AI data centres consume enormous amounts of energy, raising urgent concerns about power management, environmental impact, and sustainability. This work explores an innovative approach that makes part of the input power as dynamics that are otherwise used for computation. It evaluates passive and active solutions, comparing their environmental impact, energy efficiency, and potential to reduce capital and management costs using power trends from global data platforms.

Biography:
Andrea Marinoni holds a research professorship with the Dept. of Computer Science and Technology, University of Cambridge, UK. From 2018 to 2025, he was a professor of environmental data science with the Dept. of Physics and Technology, UiT The Arctic University of Norway, Tromsø, Norway. In 2023, he was a visiting professor at the Earth Observatory of Singapore, Asian School of Environment, Nanyang Technological University, Singapore. His main goal is to improve the efficiency and impact of AI on environmental modeling and energy-aware computation, with new strategies for data processing.