UIA – AI INFRASTRUCTURE UNIT: QUANTIFYING THE PHYSICAL CONSUMPTION OF ELECTRICITY AND WATER IN ARTIFICIAL INTELLIGENCE APPLICATIONS
DOI:
https://doi.org/10.56238/ERR01v10n7-002Keywords:
Artificial Intelligence, Digital Sustainability, Computational Efficiency, UIA, Energy Consumption, Water Footprint, Green AI, Computational InfrastructureAbstract
The accelerated growth of Artificial Intelligence (AI) has significantly increased the demand for physical resources, such as electricity and water, often neglected in conventional economic and technological analyses. This article proposes the AI Infrastructure Unit (UIA), a standardized quantitative metric designed to express the combined consumption of energy (kWh) and water (liters) per inference operation in AI models. The methodology integrates empirical data from data centers, international efficiency standards (PUE and WUE), and economic-environmental coefficients to construct a composite unit of physical cost. UIA values were estimated and compared for different large-scale language models, covering short and long query scenarios. The results show significant variations in physical consumption between models, indicating that differences in infrastructure and hardware can multiply the environmental cost per operation by up to 80 times. The analysis demonstrates that marginal efficiency gains, when applied on a large scale, represent significant reductions in energy and water. It is concluded that the adoption of UIA can contribute to environmental transparency, fair pricing, and incentives for computational efficiency. The metric proposes a new way to measure the sustainability of AI, making visible the material basis of systems that, until then, were treated as purely digital.
Downloads
References
[1] Li, P. et al. (2023). Making AI Less ”Thirsty”: Uncovering and Addressing the Secret Water Footprint of AI Models. arXiv:2304.03271.
[2] Patterson, D. et al. (2021). Carbon Emissions and Large Neural Network Training. Google Research Technical Report.
[3] Strubell, E., Ganesh, A., McCallum, A. (2019). Energy and Policy Considerations for Deep Learning in NLP. ACL.
[4] Masanet, E. et al. (2020). Recalibrating Global Data Center Energy-Use Estimates. Science.
[5] ISO/IEC 30134-2:2016. Data centres — Key performance indicators — Part 2: Power usage effectiveness (PUE).
[6] ISO/IEC 30134-9:2022. Data centres — Key performance indicators — Part 9: Water usage effectiveness (WUE).
[7] The Green Grid. (2011). White Paper 35: Water Usage Effectiveness (WUE): Usage Guidelines for the Data Center Industry.
[8] The Green Grid. (2010). White Paper 32: Carbon Usage Effectiveness (CUE): A Green Grid Data Center Sustainability Metric.
[9] Henderson, P. et al. (2020). Towards the Systematic Reporting of the Energy and Carbon Footprints of Machine Learning. JMLR, 21.
[10] Mytton, D. (2021). Data Centre Water Consumption. npj Clean Water, 4(11).
[11] International Energy Agency / 4E (2025). Data Centre Energy Use: Critical Review of Models and Results.
[12] Uptime Institute. (2025). Water is Local: Generalities Do Not Apply.
