The heat is starting to break, but plants are still wilted, and we are still conscious that water is precious — especially in California and other states in the West. Meanwhile, data centers here are sprouting like mushrooms after the rain.
And data centers require a lot of water.
In Silicon Valley, more than 55 data centers operate in the city of Santa Clara alone, and more have already been approved, according to the city. As generative AI floods into multiple aspects of our lives (work, health care, education, entertainment, access to information, companionship, national security, etc.), the need for powerful data centers grows.
Some are cooled with air; some use recycled water; many, however, require drinking-quality water for cooling. (And cooling is only part of the demand; ultra-pure water, for example, is needed for the manufacturing of the chips used in the servers running in data centers, and, according to the Environmental and Energy Study Institute, “[c]reating ultrapure water is a highly water-intensive process” itself.)
A 2024 report from the Lawrence Berkeley National Laboratory noted that in 2014 data centers “consumed 21.2 billion liters of water,” and that, in 2028, data centers optimized for AI are anticipated to require “between 60 and 124 billion liters.”
In some parts of the world, the rising demand for water for data centers has clashed with the demand for drinking water for people. Protests and boycotts related to data center water consumption have taken place in India, Uruguay, Chile, and Spain, among other places. In the Spanish region of Aragon, one activist collective calls itself “Tu Nube Seca Mi Rio”: “Your Cloud Is Drying My River.”
Several technology companies have released environmental reports highlighting their efforts to limit water consumption in their data centers, as well as various water replenishment projects. Most such reports, however, lack the kind of specificity that would allow for real understanding and planning by regulatory agencies, researchers and members of communities directly impacted by the companies’ water usage.
While clean drinking water is one of the conditions required for the common good, its use by data centers also bring up issues of environmental justice. As researchers have pointed out, different communities that host data centers are shouldering different types and levels of costs. As an initial step, environmental justice calls for accurate and much more granular disclosures related to each data center — of the type of cooling system involved, of the amount of water consumed, and of the sources of the water used.
All residents, and even innovators hoping to build better tech, are in need of better data about water use by data centers. In the EU, data center operators are required to report such information.
In California, for now, they are not.
Researchers have been calling for such disclosure for years now, with limited impact so far. While media outlets are increasingly highlighting the environmental impact of AI, including its water demands, many people (including developers and heavy users of AI tools, as well as regulators) remain unaware of this key ethical issue.
Meanwhile, the federal government is pushing for more and faster building of data centers. The White House “AI Action Plan,” for example, directs federal agencies with “significant land portfolios” to “identify sites suited to large-scale development” of data centers and related infrastructure, and calling for more relaxed permitting processes.
As the heat outside rises, and usage of generative AI continues to spread, we need to generate widespread awareness.
We need a public push for more data and more education about the water-related impacts of data centers. The speed of deployment should be matched by the speed of disclosure.
Irina Raicu is the director of the Internet Ethics program at the Markkula Center for Applied Ethics at Santa Clara University.