Ar. Kiranjeet Kaur Jassal,
Dhruvjeet Singh,
- 1Director (Sustainability Consultant), Rising Boxes Technology Solutions, Rajpura, India
- Student, Chitkara University, Rajpura, India
Abstract
Generative Artificial Intelligence (GenAI) has advanced rapidly in scale and complexity, enabling powerful capabilities in automated content creation, multimodal reasoning and real-time decision support across sectors. While these systems offer significant technological and economic benefits, their environmental implications are not fully examined. Large-scale GenAI models rely on high-performance computing infrastructure that consumes substantial energy and resources throughout their lifecycle, raising critical sustainability concerns. This paper offers a sustainability-oriented assessment of the environmental footprint of GenAI systems, focusing on energy consumption, carbon emissions, freshwater usage and material impacts. It synthesizes recent research on the environmental intensity of model training, inference workloads and data-center operations, while also considering the embodied carbon associated with hardware manufacturing and infrastructure development. The analysis demonstrates that the environmental cost of GenAI extends well beyond electricity use, encompassing significant freshwater withdrawal for cooling processes, land-use impacts from data-center expansion and increasing volumes of electronic waste driven by short hardware lifecycles. A key contribution of the study is its examination of the water–energy–carbon nexus in AI infrastructure, highlighting how GenAI deployments can intensify resource stress in climate-vulnerable and water-scarce regions. The paper further evaluates emerging mitigation strategies, including renewable-powered training, carbon-aware workload scheduling, efficient model architectures, alternative cooling technologies and circular approaches to hardware design and reuse. While these strategies offer meaningful reductions in environmental impact, the findings suggest that technical measures alone are insufficient. The study concludes that addressing the green cost of GenAI requires integrated policy frameworks, transparent sustainability reporting and a fundamental reconsideration of digital infrastructure design. By consolidating energy, water, carbon and material considerations into a unified assessment, this paper provides a valuable reference for sustainability researchers, environmental planners and organizations deploying high-performance AI systems.
Keywords: Carbon emissions, data centers, generative AI, sustainability, water footprint
[This article belongs to International Journal of Sustainability ]
Ar. Kiranjeet Kaur Jassal, Dhruvjeet Singh. The Green Cost of Generative Ai: Environmental Sustainability Implications of Large-Scale Ai Systems. International Journal of Sustainability. 2026; 03(01):12-20.
Ar. Kiranjeet Kaur Jassal, Dhruvjeet Singh. The Green Cost of Generative Ai: Environmental Sustainability Implications of Large-Scale Ai Systems. International Journal of Sustainability. 2026; 03(01):12-20. Available from: https://journals.stmjournals.com/ijsu/article=2026/view=243270
References
1. Strubell E, Ganesh A, McCallum A. Energy and policy considerations for deep learning. ACL. 2019;364–73.
2. Patterson D, Gonzalez J, Le Q, et al. Carbon emissions and large neural network training. arXiv. 2021.
3. Thompson N, Greenewald K, Lee K. The computational limits of deep learning. Commun ACM. 2022;65(8):107–15.
4. Wu X, Liu J, Li Y. Water footprint of data centres supporting artificial intelligence. Resour Conserv Recycl. 2023;185:106–25.
5. Li X, Ren Y, Zhao H. Cooling systems in high-density data centres. Appl Energy. 2022;311:118–63.
6. Rahman M, Zhang Y. Environmental lifecycle impacts of GPU computing. J Clean Prod. 2021;314:127–45.
7. Henderson P, Hu J. Reporting energy and carbon footprints of ML models. JMLR. 2022;23:1–45.
8. Gupta A, Das S. Policy frameworks for sustainable AI infrastructure. AI Ethics. 2023;4(3):222–34.
9. Balaji V, Malik IC. Carbon-aware AI scheduling. Sustain Comput. 2021;30:100–45.
10. Wang Y, Chen H. Efficient transformer architectures for low-carbon AI. Neurocomputing. 2023;540:125–40.
11. Nori S, Raji ID. Environmental transparency in AI development. Patterns. 2022;3(12):100–67.
12. Cho J, Park E. Renewable-powered data centres for emerging AI. Renew Energy. 2023;205:611–23.
International Journal of Sustainability
| Volume | 03 |
| Issue | 01 |
| Received | 05/12/2025 |
| Accepted | 20/12/2025 |
| Published | 09/05/2026 |
| Publication Time | 155 Days |
Login
PlumX Metrics