Energy-Efficient AI: Models, Algorithms, and Hardware for Sustainable Intelligence

As artificial intelligence systems grow in scale and capability, their energy demands are increasing at an unsustainable pace. This workshop addresses the urgent need for energy-efficient AI, a field that seeks to develop environmentally responsible and computationally efficient approaches to machine learning. The goal is to explore innovations that reduce energy consumption across the AI stack — from models and learning algorithms to hardware implementations — while maintaining or even enhancing performance. One source of inspiration for reaching this goal is the brain, which manages to provide intelligence with 20W.

The workshop will be inclusive and accessible to a broad audience across machine learning, computer architecture, and systems engineering. Topics include but are not limited to:

• Model compression and pruning techniques that reduce inference and training costs
• Quantization and low-precision arithmetic for efficient deployment
• New learning algorithms designed for reduced computational and memory overhead
• Architectures and model designs tailored for energy-aware operation
• Neuromorphic computing and spiking neural networks inspired by biological efficiency
• Hardware accelerators, including FPGAs, photonic processors, and neuromorphic chips, that offer novel energy-performance trade-offs

This workshop aims to bring together researchers and practitioners from diverse backgrounds to foster collaboration and cross-pollination of ideas in making AI more sustainable without compromising its transformative potential.

Organizers: Wolfgang Maass, Marcel van Gerven

In case of questions contact: Randi Goertz

Workshop Agenda

The ELLIS UnConference workshop is co-located with EurIPS Copenhagen and will take place on December 2, 2025 at the Bella Center Copenhagen, Center Blvd. 5, 2300 København S, Denmark. Our workshop will take place in room 20.

Accepted Posters