AI's water footprint extends beyond data centers
This opinion piece argues that the environmental impact of AI's water footprint extends beyond data center cooling to include water used by power plants and in semiconductor manufacturing. It highlights the OpenAI and SoftBank PORTS-Pike project in Ohio as an example, noting its indirect water consumption for Nvidia chip production. The article calls for a reevaluation of water use by advanced computing facilities.
An opinion piece from The News Record argues that the environmental scrutiny of AI-powered data centers, particularly regarding water consumption, is often too narrowly focused. While direct cooling water usage is discussed, the article asserts that the larger environmental footprint includes water consumed by the electrical grids powering these facilities and the immense quantities of ultrapure water needed for semiconductor manufacturing.
The author cites an incident involving a data center in New Jersey that resulted in an oil spill into the Hackensack River, highlighting other resource demands. The article distinguishes between open-loop (evaporative) and closed-loop liquid cooling, noting that while closed-loop systems reduce a data center's direct water use by up to 70%, the power plants supplying electricity to these facilities still consume significant water.
A key example is the OpenAI and SoftBank's PORTS-Pike project in Ohio, touted as a future world's largest data center. Despite using closed-loop cooling, the project's partnership with Nvidia will necessitate a record number of semiconductors, leading to thousands of gallons of freshwater indirectly used and contaminated during their manufacturing. The article concludes by emphasizing the need for data centers to address their indirect water consumption to prevent exacerbating water scarcity issues for communities.