Authors: Chathura S. Jayasinghe
Abstract: The rapid expansion of artificial intelligence (AI) applications has significantly increased global data center workloads, leading to rising energy demands and associated carbon emissions. Sustainable AI strategies are emerging as critical solutions to counteract these environmental challenges. This review examines the integration of AI-driven methodologies that promote sustainability within data center operations. It explores how AI can optimize energy use, reduce carbon footprint, and enable green computing practices through intelligent workload management, predictive cooling, and hardware efficiency improvements. The paper presents a synthesis of literature on carbon-aware computing, highlighting approaches such as AI-based energy forecasting, model optimization, and renewable energy integration. Furthermore, it evaluates case studies from industry leaders like Google and Microsoft, demonstrating quantifiable reductions in power usage effectiveness (PUE) and carbon usage effectiveness (CUE). Analytical frameworks and sustainability metrics are discussed to assess environmental performance, along with limitations such as data transparency and scalability challenges. Finally, the paper identifies future research opportunities in low-energy AI model development, federated learning for energy optimization, and policy-driven sustainability governance. The findings suggest that sustainable AI strategies can substantially mitigate the ecological footprint of modern computing infrastructures while ensuring computational resilience and efficiency. Through an interdisciplinary perspective, this review underscores the necessity of embedding sustainability principles into AI system design, operation, and lifecycle management. The collective insights affirm that the synergy between AI innovation and environmental responsibility is pivotal to achieving a carbon-neutral data center ecosystem.
DOI: http://doi.org/10.5281/zenodo.17882399
