AI data centres to drive global water usage 11-fold by 2028: Morgan Stanley
Cooling, electricity generation and chip manufacturing to push up consumption; over half of key hubs already face medium to high water stress.
ADVERTISEMENT
Artificial Intelligence (AI) data centres are set to drive a sharp surge in global water consumption in the coming years, according to a new report by Morgan Stanley.
The report estimates that annual water use for cooling and electricity generation by AI data centres could reach around 1,068 billion litres by 2028, an 11-fold increase from 2024 levels.
While direct water consumption for cooling is widely recognised, the report noted that the indirect water footprint through electricity generation often remains underappreciated. Semiconductor manufacturing, another pillar of the AI ecosystem, adds further pressure, with some facilities consuming up to five million gallons of ultrapure water daily.
Morgan Stanley modelled three scenarios based on water intensity, cooling technologies, and regional energy mixes, suggesting annual water consumption could range between 637 billion litres and 1,485 billion litres by 2028.
The report breaks down AI’s water footprint into three scopes: on-site data centre cooling (scope 1), electricity generation (scope 2), and semiconductor production (scope 3). Of these, electricity generation typically accounts for the largest share.
Crucially, the environmental impact is not evenly spread. More than half of the world’s top data centre hubs are located in regions already grappling with water scarcity, drought, flooding, or declining water quality, making efficient water resource management a critical challenge for the industry’s sustainability.
“Water consumption is a rapidly evolving issue for the AI sector,” the report cautioned, underscoring the need for sustainable solutions as the industry scales globally.
