Laleh Behjat

University of Calgary, Canada

Sustainability Across the AI Hardware Life Cycle

Abstract

The environmental impact of AI hardware extends far beyond operational energy consumption, encompassing material extraction, semiconductor fabrication, packaging, deployment, and end-of-life disposal. Despite growing interest in sustainable computing, life-cycle assessment (LCA) remains largely disconnected from mainstream chip design methodologies, where optimization continues to focus primarily on performance, power, and area.

This work explores how LCA concepts can be incorporated into chip design workflows. We discuss opportunities for integrating environmental metrics—including embodied carbon, water consumption, material usage, and manufacturing resource demands—into existing Electronic Design Automation (EDA) processes and system-level design decisions.

We further examine the challenges associated with embedding LCA into semiconductor design, including data availability, cross-layer modeling complexity, rapidly evolving fabrication technologies, and trade-offs between operational efficiency and manufacturing impact. By positioning sustainability as a design consideration rather than a post-production evaluation step, this work aims to support more environmentally informed approaches to next-generation AI hardware development.