While much of the AI debate focuses on its economic potential, its expanding physical footprint tells a different story. The machines driving this revolution depend on a resource far older – and far more contested – than data or electricity.
BERLIN/SANTIAGO – AI is often portrayed as the harbinger of a prosperous, more efficient future. But the machines driving this revolution depend on a resource far older – and far more contested – than data or electricity: water.
As the Heinrich Böll Foundation’s recent Water Atlas makes clear, AI’s rapid growth is depleting local water reserves around the world, from drought-stricken Chile to South Africa. Its physical footprint reflects a new form of colonial extraction; instead of silver and soy, now it is the cooling water that keeps the digital economy running. While the debate about AI’s energy use focuses on the power needed to train and operate large language models, what is often overlooked is the vast amount of water required to cool data centers, not to mention the water used in energy production and hardware manufacturing. ChatGPT is a prime example. Training GPT-3 required roughly 700,000 liters of water for cooling alone. A Greenpeace study estimates that data centers will consume 664 billion liters annually by 2030, compared to 239 billion liters in 2024. AI’s benefits are concentrated in the Global North, yet its environmental costs increasingly fall on the Global South. In 2023, mass protests erupted in Uruguay over a proposed Google data center as the country suffered its worst drought in 70 years. With reservoirs running dry, authorities began pumping brackish water from the Río de la Plata estuary into public systems, granting Google permits to draw from the remaining freshwater reserves even as working-class families boiled salty tap water to drink.
A similar conflict has unfolded in Chile, one of Latin America’s most drought-prone countries. In Santiago’s Cerrillos district, a proposed Google data center was projected to consume 7.6 million liters of water per day – roughly equal to the entire community’s annual use. In response, activists from the local group MOSACAT launched a legal and political campaign that forced a redesign of the cooling system and a new environmental review.
These community struggles highlight a familiar pattern in which corporations and governments present data centers as engines of modernization while downplaying their environmental costs. In Mexico’s Querétaro region, where rural and indigenous communities already face severe water scarcity, the problems go far beyond depletion: diesel emissions from backup generators cause air and noise pollution; electronic waste imported from the Global North continues to pile up; and growing demand for land, housing, and electricity is driving up costs and straining local infrastructure.
Regulation has done little to slow this expansion or improve environmental standards. While the European Union’s 2024 AI Act mandates transparency on energy demand and computing power, it says nothing about water use. Even the Energy Efficiency Directive, which requires data centers to report water consumption, applies only to data facilities within the EU. Moreover, reporting is not the same as reform: efficiency – limited by technology and the Jevons paradox (which occurs when greater efficiency boosts demand for a resource) – too often distracts from the deeper question of sufficiency. At the same time, many developing economies compete for tech investment by offering generous tax breaks and fast-tracking environmental permits with minimal oversight. Governments tend to frame this as advancing data sovereignty, but Big Tech ultimately holds the power. Moreover, contrary to official promises, data centers create few jobs, and structural inequalities continue to impede the growth of local AI industries. For example, criticism of Brazil’s data center policy highlights its focus on attracting large tech firms, while neglecting fair competition for domestic companies. Environmental impact assessments are another weak link. Studies show they are frequently incomplete, inaccurate, or hidden from public scrutiny.
In Chile, regulators approved Google’s project despite unresolved issues concerning groundwater rights. In Mexico, activists spent months fighting for access to water-use documents. And in South Africa and Brazil, companies often negotiate directly with national ministries, bypassing local authorities altogether. All of this raises a critical question: Who gets a say when digital growth depends on local water sources? Like its benefits, the risks of AI are unequally distributed. For many Latin American and African communities, opposition to data centers is not a rejection of progress but an effort to redefine it. Their defense of water reserves challenges the fantasy of infinite digital expansion in a world of finite resources.
The problem isn’t one of innovation but of distribution. Sustainable cooling systems that use recycled water, saline water, and rainwater already exist, and air-based systems and heat recovery can further reduce freshwater use. But companies have little incentive to adopt these alternatives when water is cheap, unregulated, and invisible on balance sheets. Another, deeper problem lies in AI’s very nature: its intensive computing demands ever-greater water consumption. Addressing these challenges requires reconciling technological ambition with the realities of today’s escalating climate and ecological crises. Otherwise, AI’s unchecked growth risks turning water-stressed regions into sacrifice zones. This task – shaping a humane and sustainable technological future – is not one that individuals and communities can accomplish on their own. Political leaders must take urgent steps to democratize decision-making, ensure accountability, and align technological innovation with planetary boundaries.
This article was first published on Project Syndicate. The views and opinions in this article do not necessarily reflect those of the Heinrich-Böll-Stiftung European Union | Global Dialogue.
