AI and Energy: Building a Sustainable Future
In collaboration with the Applied Machine Learning Days, we will explore holistic approaches to sustainable computing, from renewable energy integration and advanced cooling in data centers to hardware efficiency with custom accelerators, energy storage solutions, heat reuse, and the broader vision of sustainable AI.
As AI demand continues to rapidly increase, its growing energy demand calls for urgent and holistic solutions to be sustainable. During this event, the talks will address sustainable computing across multiple layers, from renewable energy integration and advance cooling technologies in data centers to energy-efficient hardware design and custom accelerators. Featured talks include pioneering initiatives such as EPFL’s HeatingBits and EMPA’s HEATWISE project, which demonstrate how data centers and edge computing facilities can smartly be integrated to be more sustainable by enhancing heat reuse and possibly reducing drastically carbon emissions, through different means such as battery storage systems and advanced heat reuse and storage.
We will also high the role of AI itself in optimizing its own footprint. Beyond technological breakthroughs, we aim to promote sustainability as a fundamental design principle for future computing systems. Together, these contributions define a roadmap toward AI that is not only powerful and intelligent but also energy-responsible and sustainable by design.
Talks:
- The energy wall: success and survival strategies for AI – Prof. Giovanni De Micheli, EcoCloud, EPFL
- The HEATWISE project: Waste heat utilization from edge data centers in tertiary buildings – Dr. Binod Koirala, Urban Energy Systems Lab, EMPA
- Improving data centers’ energy efficiency while reducing their carbon footprint: the EPFL HeatingBits project – Prof. Drazen Dujic, Power Electronics Laboratory, EPFL
- Beyond the Bottleneck: AI as the catalyst for sustainable computing – Dr. Darong Huang, Embedded System Laboratory, EPFL
Chair: Dr. Xavier Ouvrard, EcoCloud, EPFL
