Sustainable Intelligence at the Edge: Using Sustainability as a Metric for AI Cloud-Edge Design

Abstract

The swift growth of Cloud-Edge infrastructure to support IoT and artificial intelligence workloads with low latency is exceeding the capacity of our management to control the environmental impact. Whereas there are energy-efficient hardware and software, the measures of sustainability, such as lifecycle assessment (LCA), carbon intensity, and circularity, are inconsistently defined and seldom reported. This piece of work brings sustainability to first-class design and operations measures of CloudEdge systems, inquiring how it can be directly incorporated into benchmarking and decision-making. The literature review and cross-venue trend analysis (2018-2025) have been conducted in a structured way to map existing practices and gaps. Based on these results, we suggest the Sustainability-conscious Hybrid Edge Assessment Blueprint (SHEAB): a workflow that integrates automated lifecycle inventory, power/CO2 estimation at workload, and circular economy into what is today taken as performance benchmarking. The review is a revelation of ongoing LCA scope, boundary setting, and reporting gaps; in its turn, SHEAB unifies a small, similar metrics portfolio, e.g., Performance-per-Watt (PPW), CO2e per inference/transaction, and embodied carbon per device, and offers a sustainable evaluation pipeline. Using the blueprint to calculate representative edge workloads and device classes shows end-to-end feasibility and uncovers viable trade-offs between throughput and carbon intensity among deployment options. When sustainability is the key metric, the choices of the system change: hardware is selected, its location is different, and the timeframe is varied when CO2e and circularity can be placed next to latency and price. We end up with reporting guidelines and lessons learned to be put into action to fast-track the truly sustainable intelligent edge systems as the AI demand and its carbon footprint continue to increase.