Why the true water footprint of AI is so elusive
Briefly

As data centers grow to meet the demands of AI, concerns about their environmental impact, particularly regarding electricity and water usage, escalate. The 2024 U.S. Data Center Energy Usage Report estimates these facilities could consume 12% of the nation’s electricity by 2028, while studies reveal alarming water consumption statistics tied to AI training. With the absence of regulatory frameworks for disclosing energy and water usage, researchers like Shaolei Ren emphasize the need for awareness and accountability in the tech industry. Tech giants have pledged to achieve net-zero carbon emissions by 2030, but questions about their water footprint linger.
"We only had water access for like half an hour each day. So we just had to use water very wisely," says Shaolei Ren, illustrating the importance of water conservation.
The 2024 United States Data Center Energy Usage Report forecasts that U.S. data centers could consume as much as 12% of the nation's electricity by 2028.
With no federal or state regulations mandating tech companies to disclose their energy and water consumption, the true environmental footprint of AI remains uncertain.
Ren was among the first to study the water footprint of AI, estimating that training the GPT-3 model consumed hundreds of thousands of liters of fresh water.
Read at www.npr.org
[
|
]