Abstract: Artificial intelligence has rapidly evolved into a critical technological driver across numerous domains including healthcare, finance, transportation, and large-scale data analytics. Recent developments in deep learning have significantly improved the capabilities of artificial intelligence systems , particularly through the adoption of large neural network architectures that require extensive computational resources for training and deployment While these advancements have enabled improved predictive performance and broader application potential they have also resulted in substantial increases in computational power requirements and associated with the energy consumption. The growing dependence on high performance computing infrastructures has raised concerns regarding the environmental sustainability of artificial intelligence technologies. Training modern deep learning models often involves the use of multiple graphics processing units (GPUs) operating for extended durations, which leads to the considerable electricity consumption and indirectly contributes to carbon emissions depending on the energy source used by computing facilities. This study analytically investigates the relationship between computational power requirements of artificial intelligence systems and their environmental impact. The analysis focuses on key computational parameters including model complexity, hardware utilization, and training duration to estimate the energy consumption associated with AI model training processes. These energy values are subsequently translated into carbon emission estimates using carbon intensity metrics to evaluate the environmental implications of AI computation. The findings reveal that increases in model scale significantly amplify both energy consumption and carbon emissions, highlighting the need for energy-efficient artificial intelligence frameworks. The study proposes the adoption of energy-aware AI development strategies and computational optimization approaches to promote sustainable artificial intelligence systems.

Keywords Artificial Intelligence, Green AI, Energy Consumption, Carbon Emissions, Sustainable Computing, Deep Learning Efficiency

Downloads: | DOI: 10.17148/IARJSET.2026.13341