Green AI framework aims to make smart cities computationally sustainable

As cities grow, the systems that keep them running—electricity grids, water networks, transport and waste management—face mounting pressure. Artificial intelligence has proven effective at optimizing these urban operations, from predicting electricity demand to routing traffic. But the very models that make cities smarter are themselves becoming energy hogs, and a new survey argues that this computational appetite threatens to undermine the sustainability goals AI is meant to serve.

Published in Artificial Intelligence Review, the study reviews recent advances in Green AI methodologies and introduces the Green AI Urban Framework, or GAUF, a structured conceptual framework for embedding environmentally responsible AI into sustainable urban systems. The authors, led by Elda Cina and colleagues, survey techniques across four fronts: energy-efficient model design, lightweight and adaptive learning, data-efficient training, and resource-aware deployment strategies. Each addresses a different stage of the AI lifecycle, from how models are built to where they run.

The survey examines major urban application domains—electricity demand prediction, water resource optimization, waste management and smart mobility—to identify current capabilities, limitations and research gaps. The analysis is a consolidation of existing literature rather than a new empirical evaluation, so it maps what has been demonstrated in prior work rather than reporting fresh performance results. Its value lies in organizing a fragmented field: green AI techniques exist, but until now there has been no unified structure for applying them specifically to urban infrastructure.

GAUF integrates three core components. The first is sustainability-aware model selection, which weighs environmental cost alongside predictive accuracy when choosing algorithms. The second covers cloud–edge deployment strategies, distributing computation between centralized data centers and local edge devices to reduce energy overhead and latency. The third establishes energy-efficiency evaluation metrics that span the entire AI lifecycle, from training through inference, so that a model’s environmental footprint can be measured rather than assumed.

Beyond the technical core, the framework incorporates implementation and governance considerations relevant to urban infrastructure, sustainability reporting and municipal adoption. This acknowledges a practical reality: cities do not simply deploy algorithms, they procure, regulate and audit them. A framework that ignores procurement rules or reporting obligations would struggle to move from paper to practice, and the authors position GAUF as a bridge between AI researchers, practitioners and policymakers.

The broader tension the paper highlights is genuine. AI’s computational complexity and energy consumption have grown alongside its capabilities, creating what the authors describe as new sustainability challenges within systems designed to solve sustainability problems. An electricity demand forecast that saves megawatt-hours on the grid may, if trained and served inefficiently, consume a meaningful share of those savings itself. Quantifying that trade-off requires the kind of lifecycle metrics GAUF proposes, which the survey identifies as a current research gap.

As a conceptual contribution, GAUF does not report measured energy savings or benchmarked deployments; its claims are about structure and direction, not demonstrated outcomes. The authors state its aim explicitly: to support the design of AI-driven urban systems that balance predictive performance, scalability and environmental responsibility. By consolidating existing research and outlining future directions, the study positions sustainable AI practice as a core component of next-generation smart and resilient cities, while leaving empirical validation of the framework to future work.

Subject of Research: A green AI framework and survey for environmentally sustainable smart city systems

Article Title: Designing sustainable urban intelligence: a green AI framework and survey

Article References: Cina, E., AlArnaout, Z., Elbasi, E., & Elmazi, G. (2026). Designing sustainable urban intelligence: a green AI framework and survey. Artificial Intelligence Review. https://doi.org/10.1007/s10462-026-11730-3

Image Credits: AI Generated

DOI: 10.1007/s10462-026-11730-3

Keywords: Green AI, smart cities, urban sustainability, energy-efficient AI, cloud-edge computing, electricity demand prediction, water resource optimization, waste management, smart mobility, AI governance, sustainability metrics, Designing