Yi Ding

As artificial intelligence and data centers expand, their hidden water costs grow.

Yi Ding, assistant professor in Purdue University’s Elmore Family School of Electrical and Computer Engineering, has been named project lead for the Software Water Intensity project, a new initiative launched within the Green Software Foundation (part of Linux Foundation). The project aims to develop a consistent way to measure the water consumed by software, similar to how organizations now measure software’s carbon footprint.

The Software Water Intensity, or SWI, project is led by the foundation’s Software Standards Working Group. Its goal is to create a standard framework that helps companies, researchers and engineers understand, report and reduce the amount of water connected to running software. By providing a consistent method for evaluating water consumption impact, the project empowers organizations to identify inefficiencies, meet sustainability goals, and communicate their software’s environmental impact transparently.

Although software itself may not directly consume water, the computers and data centers that run it do. Data centers, particularly those supporting AI systems, use water for cooling. Additionally, water is consumed indirectly through electricity generation and in the manufacturing of semiconductors, the chips essential to modern computing. Depending on design, location, and workload, large data centers may consume millions of liters of water each day.

Despite this significant impact, there is currently no shared industry standard for measuring the amount of water associated with running software.

"The scale at which AI and data centers are consuming water is unprecedented, yet the industry lacks a consistent measurement methodology," Ding said. "Developing the Software Water Intensity specification addresses this gap. I am honored to lead this project and to collaborate with researchers and industry practitioners who believe that water-aware software is both necessary and achievable."

The project builds on the Software Carbon Intensity specification, which helped make software’s carbon emissions easier to measure and report. The Green Software Foundation says a similar approach is needed for water as companies face growing sustainability reporting requirements and as more data center operators begin sharing water-use data.

The SWI project is designed for cloud and infrastructure providers, hardware and semiconductor manufacturers, AI platform companies, and organizations that need a credible way to measure software’s water footprint. By using the SWI framework, these groups can make more informed decisions to improve water efficiency, adhere more closely to regulatory requirements, and enhance their sustainability reporting.

At Purdue, Ding leads the STYLE Lab, short for Sustainable and Trustworthy Computing Systems and Learning. Her research focuses on AI and machine learning systems and the environmental impacts of computing, including carbon, water, and biodiversity.

The SWI project is now in incubation and open to contributions from Green Software Foundation member organizations.