Generative Artificial Intelligence (GenAI) is frequently hailed as a revolutionary force, yet its rapid adoption often overlooks the substantial environmental costs hidden behind the metaphor of the cloud. This paper critically examines the physical infrastructure of GenAI, arguing that the current trajectory represents a "gluttonous fraud" of resource consumption compared to the "genuine intelligence" of sustainable computing. Through a synthesis of recent environmental impact studies, this research deconstructs the four pillars of AI’s footprint: energy consumption dominated by inference rather than training, the hidden water usage for cooling and electricity generation, the material toll of e-waste and mineral extraction, and the lack of corporate transparency. The analysis reveals that the efficiency gains of current models paradoxically drive higher consumption and that specific tools actively propagate non-green code. The study concludes that without intervention, GenAI will exacerbate the climate crisis through a "computational opulence paradox." Consequently, it proposes actionable pathways toward Green AI, including the adoption of smaller, specialized models, mandatory environmental impact reporting, and regulatory frameworks. These findings imply a necessary paradigm shift for policymakers and academic leaders from unfettered adoption to a model of "slow urgency" and rigorous accountability.
Bozkurt et al. (Thu,) studied this question.