Artificial Intelligence Wants Our Water
Artificial intelligence is often portrayed as a harbinger of a more prosperous and efficient future. However, the machines driving this revolution rely on a resource that is much older—and more contested—than data or electricity: water.
As highlighted in the recently released Water Atlas by the Heinrich Böll Foundation, the rapid growth of artificial intelligence is draining local water reserves worldwide, from drought-stricken Chile to South Africa. Its physical footprint reflects a new form of colonial extraction; instead of silver or soy, cooling water has now become essential for the digital economy’s sustainability.
While the debate around AI’s energy usage focuses on the amount of power needed to train and operate large language models, the vast quantities of water required for cooling data centers often go unmentioned, not to mention the water used in energy production and device manufacturing.
ChatGPT serves as a prime example. Training the massive language model GPT-3 alone required nearly 700,000 liters of water just for cooling. Estimates by Greenpeace suggest that data centers will consume 664 billion liters annually by 2030, up from approximately 239 billion liters in 2024.
The benefits of AI are concentrated in the Global North, yet its environmental costs increasingly burden the Global South. In 2023, massive protests erupted in Uruguay over a proposed Google data center while the country faced its worst drought in 70 years. As stocks dwindled, authorities began to pump slightly saline water from the Río de la Plata estuary into public networks, while Google was granted permits to draw from the remaining freshwater reserves as working-class families boiled salty tap water for drinking.
A similar conflict unfolded in Chile, one of Latin America’s most drought-prone countries. In the Cerro Los Patrimonios area of Santiago, a proposed Google data center was expected to consume 7.6 million liters of water daily—equivalent to nearly the annual usage of the entire community. In response, activists from the local MOSACAT group launched a legal and political campaign that forced a redesign of the cooling system and a new environmental review.
These community struggles underscore a familiar narrative in which corporations and governments present data centers as engines of modernization while downplaying their immense environmental costs. In Querétaro, Mexico, rural and Indigenous communities are already experiencing severe water scarcity. The problems extend beyond depletion: diesel emissions from backup generators pollute the air, and imported electronic waste from the Global North continues to accumulate; increasing demands for land, housing, and electricity raise costs and stress local infrastructure.
Regulatory controls have done little to slow this expansion or improve environmental standards. While the European Union’s 2024 Artificial Intelligence Act mandates transparency in energy and computing power needs, it lacks provisions regarding water usage. Even the Energy Efficiency Directive, which requires data centers to report water consumption, applies only to facilities within the EU. Reporting, moreover, is not the same as reform: efficiency—constrained by technology and Jevons Paradox (the phenomenon where increased resource efficiency leads to higher demand)—often distracts from the deeper issue of sufficiency.
At the same time, many developing economies compete for investment in technology by offering generous tax breaks and fast-tracking environmental permits with minimal oversight. Governments tend to frame this as a boost to data sovereignty, but it is ultimately the tech giants that wield the power. Moreover, contrary to official promises, data centers create few jobs, and structural disparities continue to hinder the growth of local AI industries. For instance, criticism of data center policies in Brazil highlights their focus on attracting large tech companies while neglecting fair competition that benefits local businesses.
Environmental impact assessments represent another weak link. Studies frequently emerge incomplete, inaccurate, or shielded from public scrutiny. In Chile, regulators approved a Google project despite outstanding issues regarding groundwater rights. In Mexico, activists fought for months to access water use documents. In South Africa and Brazil, companies often negotiate directly with national ministries, bypassing local authorities entirely.
All of this raises a crucial question: who has a say when digital growth relies on local water sources? Just as the benefits of AI are unevenly distributed, so too are its risks. For many communities in Latin America and Africa, opposing data centers is not a rejection of progress but an attempt to redefine it. Defending water reserves challenges the imagination of boundless digital expansion in a world with finite resources.
The issue is not one of creativity but of distribution. Sustainable cooling systems that utilize recycled water, saline water, and rainwater already exist, and air-based and heat recovery systems could further reduce the consumption of freshwater. However, companies lack significant incentives to adopt such alternatives when water is cheap, unregulated, and invisible in budget sheets. A more profound problem lies in the nature of AI itself: its intensive computational processes require increasing amounts of water.
Addressing these challenges demands reconciling technological ambition with the reality of today’s escalating climate and environmental crises. Otherwise, unchecked AI growth threatens to transform water-scarce regions into sacrifice zones.
This task—shaping a humane and sustainable technological future—is not one that groups of individuals or communities can accomplish alone. Political leaders must take urgent steps to democratize decision-making processes, ensure accountability, and align technological innovation with planetary boundaries.
