Introducing Our New Project: Rooted Clouds

It is increasingly well-documented that the expansion of Artificial Intelligence (AI) is fueling a rapid proliferation of data centres worldwide. These facilities, essential for storing, processing, and training AI models, are notoriously resource-intensive. They consume vast quantities of electricity and require significant amounts of water for cooling–especially in hyperscale installations. For instance, a 2024 Washington Post investigation revealed that training ChatGPT‑4 consumes enough energy to power 14 LED light bulbs for an hour per 100‑word response, and about 500 mL of water.

Microsoft reported that its data centres used nearly 1.7 billion gallons of water in 2022–an increase linked in part to its AI operations, as noted in this Associated Press article. Google also saw a 20% jump in water usage in the same period, reaching over 5 billion gallons globally. In parts of Northern Virginia’s "data centre alley", data hubs from Amazon, Google, and Microsoft consumed around 1.85 billion gallons of water in 2023, according to the Financial Times.

These pressures on natural resources are triggering growing resistance in drought-prone communities like Arizona, Chile, and the Netherlands, where residents and activists raise concerns about water access, environmental degradation, and democratic oversight of permitting processes, as described by Business Insider. The stress is particularly acute in rural areas, where power grids and aquifers are less resilient.

While several initiatives are underway to monitor different aspects of digital infrastructure, their scopes are often either too narrow or too broad to capture the full picture of AI’s environmental impacts. For instance, platforms like Data Center Map focus on mapping the physical locations of data centers globally, offering a technical overview but little insight into their social or ecological consequences. On the other end of the spectrum, the Environmental Justice Atlas compiles case studies of environmental conflicts and injustices worldwide, many of which involve extractive industries or large-scale infrastructure, but rarely focus on the digital sector specifically. Our goal is to do something in between: to map data centers not only by their location, but also by their planetary justice impacts–including resource use, local opposition, and geopolitical patterns–thereby filling a critical gap in both environmental justice and AI infrastructure monitoring.

That’s why we are excited to launch "Rooted Clouds: Mapping AI Data Centers", a project born from the need to connect the dots and ground the AI cloud in its physical realities. Our goal is to uncover, wherever possible, key aspects such as: where these  danta centres are located, who owns them, how much water and energy they consume, who or what they impact, and what their presence means for the territories that host them.

In the coming months, we’ll be sharing updates, watch/read lists, and other resources to help build a collective narrative that advances eco-social accountability in the development of AI. From the AIPJ Team, Nicolás and Lakshmee will be leading this effort.

 

From Nicolás

Back in 2022, when I was writing my master’s thesis on the relationship between land defenders and technology in Colombia, I already knew the themes I cared about: data colonialism, environmental justice, extractivism, inequality. But they were loosely scattered, without a clear path. I didn’t yet have a project. Just a persistent sense that there was a lot of work to be done on issues.

Years later, while researching The Data Centre Dilemma, an article about the environmental impact of AI infrastructure, I came across the AI + Planetary Justice Alliance. Bit by bit, the puzzle not only started coming together, it began to expand. The Alliance’s AI Supply Chain Impact Framework made me realize just how deep the problem goes, and how many gaps remain in understanding AI’s supply chain and environmental costs.

 

From Lakshmee

My interest in AI data centers’ material impacts was seeded when a conversation with my mother revealed the mounting groundwater crisis in my hometown, Bangalore. Considered the ‘Silicon Valley of India’ given the concentration of tech companies, Bangalore is a prime location for data center buildout. Bangalore has also faced chronic water scarcity, and recently faced  its worst water crisis in 500 years. I was initially motivated to investigate the potential linkages between groundwater depletion and data center buildout. 

Now, I am keen to explore, unpack, and amplify the wider set of climate impacts that AI stands to pose, and surface timely mitigation pathways. The research on data centres fit with the work of the Alliance. It was the start of something bigger that is beginning to take shape.

 

Why “Rooted Clouds”?

Because behind every AI model, chatbot, or image generator is a material world: servers, cables, cooling systems, power plants. Land. Water. Labor.

This project is based on an urgent premise: if we are to take AI seriously, we have to trace its planetary justice costs. We need to know where this infrastructure sits, how it operates, who owns it and who is impacted. Then, we must collectively and collaboratively build uniform yet context relevant measures to mitigate these impacts. 

What We’re Doing

Together with a small team and emerging collaborators, we’re building a global and regional map of major AI data centers. Not just their locations, but their planetary justice footprints–things such as water use, energy sources, emissions, and land occupation. This will be done using the AI Supply Chain Impact Framework as the starting point of what we want to inquire about. At least, as far as we can get.

While this is a global mapping effort, our focus is grounded in place. We’re especially interested in local examples, where the impacts of these infrastructures are felt most directly. We want to understand how AI data centers reshape specific territories, affect local ecosystems, and intersect with the lives of nearby communities. That means going beyond numbers and technical specs to ask: Who lives next to these centers? What do they know? What have they been told? What are they resisting (if they are)? 

We’re also examining the social consequences for surrounding communities, the promises made around job creation, the deals cut between governments and tech giants, and the resistance already emerging in places like Chile, Spain, the Netherlands, and North America.

This is not just about surveillance capitalism, carbon emissions or AI Ethics. It’s about accountability, voice, and power. And what we’re trying to do is gather the existing efforts, both inside and outside the margins, and set them on a shared path. While a tangible aim of this work is to develop a robust and grounded evidence base to prompt timely action, the ultimate goal is to build broader public awareness of the material costs of the AI age. 

Who This Is For

We’re envisioning this not as a closed project, but as an open invitation. The more perspectives involved, the more complete the research and the mapping tool will be. This work is for researchers who want to ground their AI ethics in the material world, for journalists investigating the hidden layers of digital infrastructure, for civil society groups organizing at the local level, for policymakers, and, importantly, for the communities directly impacted by these infrastructures. 

We’re currently building the first version of the database, pulling confirmed cases like Meta’s campus in Talavera de la Reina, AWS facilities in Aragón, Google in Chile, and Microsoft projects across the U.S. and Europe.

The plan is to first launch a pilot map , followed by short features amplifying the voices of people on the frontlines: activists, researchers, communities. Eventually, we hope this will become an open-access platform where information can be visualized, tracked, and used as a tool for advocacy and analysis.

What’s Next

This blog is the first of many. A kind of collaborative research diary, documenting not just what we’re finding, but how we’re approaching the work: the hopes, obstacles, failures, and successes.

In the coming weeks, we’ll be developing the analytical framework that will guide the mapping. That means defining our indicators and identifying which data sources are viable, transparent, and accessible. 

This is an early step in a longer process. The questions we’re trying to answer aren’t easy, but ignoring them would only deepen the planetary challenges we’re already facing. If you’re working on similar issues or have knowledge to share, we’d like to hear from you. You can reach us at nicolas@aiplanetaryjustice.com.

More soon.
Nicolás and Lakshmee