Research

Generative artificial intelligence (AI) models have risen to prominence, but their development and deployment carry significant environmental implications. This paper examines the sustainability of generative AI models in cloud computing infrastructure through a comprehensive life cycle assessment (LCA) and an analysis grounded in environmental ethics. Using only peer-reviewed and verifiable data, we quantify the energy usage and greenhouse gas emissions associated with training and running large AI models, drawing on reported metrics from major cloud providers (Amazon Web Services, Microsoft Azure, Google Cloud, IBM Cloud) and recent academic studies. Our findings indicate that training a single state-of-the-art generative model can consume on the order of 1–1.3 GWh of electricity, emitting hundreds of tonnes of CO2 [1], while the widespread inference use of such models can multiply this footprint manyfold [2]. Cloud data centers offer efficiency advantages and increasingly utilize renewable energy, mitigating some impacts [3] [4]. However, rapid growth in generative AI demand is outpacing these sustainability gains, raising ethical concerns about resource consumption and climate impact. In response, we propose a framework of environmental ethics for AI, emphasizing principles of transparency, responsibility, and stewardship. We discuss strategies for reducing the life-cycle carbon footprint of AI models, including improvements in hardware efficiency, algorithmic optimizations (“Green AI”), carbon-aware computing practices, and stronger governance. The study concludes that aligning AI development with sustainability goals is both an ethical imperative and a feasible pursuit, requiring concerted efforts from researchers, industry, and policymakers. Each section of this manuscript transitions to the next by linking technical findings with broader implications, ensuring a cohesive narrative from quantitative assessment to ethical discussion.

DevOps has emerged as a transformative approach to software delivery, integrating development and operations through automation, collaboration, and continuous improvement. Despite widespread adoption, the empirical validation of DevOps maturity models, practices, and frameworks remains fragmented. This systematic literature review (SLR) examines 27 high-quality studies published between 2017 and 2025 to synthesize evidence on DevOps practices, empirical validation gaps, and adoption challenges. Following a rigorous PRISMA-based protocol, we searched five digital libraries (ACM Digital Library, IEEE Xplore, Scopus, SpringerLink, arXiv) and applied strict quality assessment criteria. Our findings reveal that CI/CD practices are nearly universal (89% of studies), while only 15% of studies employ rigorous longitudinal or controlled empirical designs, highlighting a critical validation gap. The most frequently reported challenges include cultural resistance, skills shortages, and tool complexity. This review provides evidence-based insights for practitioners and identifies critical research gaps requiring longitudinal validation, standardized metrics, and cross-context generalizability studies.