Generative AI Data Centers and the Net-Zero Challenge: A Multi-Dimensional Assessment of Energy Demand and Carbon Emissions
Main Article Content
Abstract
As Generative Artificial Intelligence (AI) is booming, demands for large-scale data center infrastructure are skyrocketing and electricity usage, carbon emissions and global Net-Zero goals are a growing concern. This study introduces a lightweight Python-based simulation framework to evaluate the long-term environmental sustainability of Generative AI data centers across four key development pathways: Baseline, High Growth, Efficiency, and Green Transition. The proposed framework combines publicly available energy and environmental data sets to provide an estimate of annual electricity demand, carbon emissions, renewable energy contribution, Power Usage Effectiveness (PUE), cumulative emissions, and the Net-Zero Gap Index (NZGI) from 2025 to 2050. Results from simulations show that significant differences in environmental performance exist between the scenarios investigated. The least sustainable development pathway is the High Growth scenario which has the highest demand for electricity (3000 TWh), carbon emissions (540 MtCO₂e), and cumulative carbon emissions (13,234.70 MtCO₂e) by 2050. The Green Transition scenario both reduces electricity demand (1350 TWh) via optimising operational performance in the system, and provides an energy mix with the lowest annual carbon emissions (64 MtCO₂e) and highest level of renewable energy (90%) to meet future climate targets, with the smallest Net-Zero Gap Index (1.88), showing greatest alignment with future climate targets. The results show the potential benefits of optimizing the use of energy in data center operations, and how quickly renewable energy can be deployed to maximize potential environmental impacts of Generative AI infrastructure. The proposed framework is transparent, computationally efficient and reproducible, and can offer decision support tools for researchers, policymakers and industry stakeholders to follow sustainable AI development paths towards Net-Zero.
Article Details

This work is licensed under a Creative Commons Attribution 4.0 International License.