New Delhi: A report by Morgan Stanley indicates that Artificial Intelligence (AI) data centres are projected to greatly increase global water consumption in the upcoming years.
According to the report, the annual water usage for cooling and electricity generation by AI data centres could reach approximately 1,068 billion litres by 2028.
"AI data centres are expected to elevate annual water consumption for cooling and electricity generation to around 1,068 billion liters by 2028 (our base case) - an 11-fold increase from estimates in 2024," it mentioned.
The report highlighted that while the water usage for cooling data centres is well recognized, the indirect water consumption through electricity generation is often overlooked. Additionally, AI's overall water impact encompasses semiconductor manufacturing.
Facilities engaged in semiconductor production can utilize up to five million gallons of ultrapure water daily, underscoring the industry's reliance on water-intensive operations.
Furthermore, the report stated that water consumption projections are influenced by various factors, including assumptions regarding water intensity, consumption rates, cooling technologies, and the regional energy mix.
As operations implement more efficient practices, these estimates may evolve. To account for this variability, the report outlined three potential scenarios. In these scenarios, AI's water usage could fluctuate between 637 billion litres and 1,485 billion litres each year by 2028.
The water footprint of AI is categorized into three scopes. Scope 1 pertains to on-site data centre cooling, scope 2 relates to electricity generation, and scope 3 involves semiconductor manufacturing.
Electricity generation (scope 2) generally represents the largest portion of water usage, followed by cooling (scope 1) and semiconductor manufacturing (scope 3).
While the overall global water consumption attributed to AI may seem relatively low, the report pointed out that its effects are significantly localized.
Over fifty percent of the foremost global data center hubs are located in areas that are already experiencing medium risks associated with water scarcity, drought, flooding, and deteriorating water quality.
In a similar vein, numerous secondary markets and emerging data center hubs are found in regions facing high or very high water stress.
The report emphasized that water consumption is an issue that is rapidly changing for the AI industry.
It noted, "More than half of the world's leading data center hubs are situated in regions that are already encountering medium basin physical risk (i.e., medium level vulnerability) due to threats from drought, flooding, and declining water quality."
As the demand for cooling and electricity generation continues to rise, the efficient management of water resources will be essential for the growth and sustainability of the industry.
