Abstract
Given the vast resources required to operate data centers today, their environmental impacts have gained increasing attention. As awareness grows, this industry is facing mounting pressure and has begun announcing mitigation strategies aimed at strengthening its environmental and social responsibility. However, are these strategies effectively addressing local environmental impacts and benefiting communities?
To answer this question, this paper proposes a ‘feral’ approach by weaving ‘feral patches’. That is to elucidate how nonhuman beings are affected by the impacts of large-scale data centers. To do so, we present two Latin American case studies situated in Quilicura (Chile) and Queretaro (Mexico). Both sites are emerging as key data center hubs exposing signs of ferality: thirsty forests and expansive droughts. This paper contends that shifting from ‘global’ or ‘planetary’ scales to ‘local’ and ‘patchy’ approaches reveals the contradiction of mitigation strategies to deliver promised benefits. Instead, these compensations are increasingly entangled in the feral effects within the Anthropocene.
Mother nature accountability
In 2021, Apple released a commercial titled Mother Nature featuring actress Octavia Spencer as a personified Mother Nature demanding accountability for the tech giant's environmental impact. During her visit, Apple executives, led by CEO Tim Cook, presented the company's extensive efforts to reduce carbon emissions, harness renewable energy, undertake environmental restoration projects, and incorporate recycled materials into their products. The commercial highlights Apple's ambitious goal of achieving carbon neutrality portraying the company as a global leader in environmental stewardship worldwide. This personification of Mother Nature serves as a powerful metaphor for the company's accountability towards the planet. As a result of emerging critiques of their environmental impact and negative social consequences, big tech companies develop narratives and marketing campaigns that communicate their environmental compensatory strategies (Hogan and Blue, 2024). In doing so, tech firms aim to address critics and wash their image regarding their environmental impacts. Aware of their increasingly poor reputation regarding environmental impacts, these firms have also been investing in social and environmental projects supposedly by bringing benefits to local communities and territories. This scenario brings the following research questions: how are big tech companies designing compensatory projects in Latin America, specifically in Chile and Mexico? How are local communities involved? How is nature reacting to these projects? And are these projects satisfactorily mitigating environmental and social impacts, or are they becoming another way of framing their business activities as environmentally and socially responsible?
Fueled by the expansion and growth of cloud computing and the projected surge of artificial intelligence (AI), hyperscale data centers have emerged as a global phenomenon (CBRE, 2024). According to the IEA, data center electricity consumption accelerated from an annual growth rate of 3% between 2005 and 2015 to 10% from 2015 to 2024 (D’Ambrosio et al. 2025). This growth trajectory, however, is intrinsically linked to an escalating demand for energy and water resources required by large-scale AI, the so-called Generative AI (Crawford, 2024; Mytton 2021; Li et al., 2023). While the size of AI models has increased over the years, this trajectory is further accelerated by the rapid growth in users. For instance, ChatGPT reached 1 million users in just five days, whereas it took Instagram over two months to reach the same number of users. This surge in users of AI-based digital services, that demand huge amounts of energy and other resources, comes with sustainability challenges.
The global surge in hyperscale data centers powering these digital services is having increasingly significant effects, including concerning environmental impacts. As van Wynsberghe (2021, p. 215) reflects ‘the AI which is being proposed to power our society cannot, through its development and use, make our society unsustainable’. Critical scholars across various disciplines are shifting their focus on the sustainability of AI, and specifically, how the AI industry is contributing to global warming rather than serving as part of the solution. Google declared that AI products have driven up their carbon emissions by 48% in 2023 compared to 2019 (Google, 2024a). To mitigate these impacts, Google has invested in carbon offset projects by purchasing carbon credits to compensate for its emissions. This trend critically demands to not only examine the environmental implications of the sustainability of AI, but also their environmental and social mitigation strategies.
In this paper, we analyse how compensatory projects are celebrated by the digital industry as solutions without a clear and local understanding of how they are addressing environmental costs or bringing social benefits in the long term. The baselines of compensatory initiatives, reference points that determine the initial conditions of the ecosystems being restored, are frequently ambiguous or entirely absent (Ureta, Lekan and von Hardenberg, 2020). This ambiguity extends to the temporal scales over which these projects are evaluated. In the short term, such initiatives frame their environmental efforts primarily through marketing discourse, downplaying their material impacts (Hogan, 2018). In the medium and long term, the sustainability and actual impact of these projects remain uncertain, for instance by showing that carbon removal projects do not actually remove carbon from the atmosphere as critical literature on political ecology demonstrates (Watt, 2021; Buck, 2022; Morrow et al. 2020). Moreover, science and technology studies have also interrogated about the sustainability of data centers by illustrating contradictions with renewable energy systems (Brodie, 2024; Laaksonen & Frig, 2025).
The contribution of this paper is a methodological analysis of the compensatory projects by data infrastructures in Latin America incorporating the ‘ferality’ framework. This approach is coined by Tsing and Bazzul (2022) recognizing the ways that living beings and non-living beings react to human infrastructural projects. In these instances, data centers operating in Latin America, along with their associated compensatory projects, such as the ‘Quilicura Urban Forest’ in Chile and the ‘Planning for a better quality of life in data center communities’ in Mexico, are neither adequately mitigating their environmental impacts nor delivering meaningful benefits to local communities. The tendency to idealize a return to a ‘pure’ state of Mother Nature fails to account for the inherent complexities and timeframes required for environmental and social compensation within the current climate crisis. In what follows, the paper presents the context, theoretical framework and empirical material gathered to investigate the feral entanglements of data infrastructures in Latin America. We argue that thirsty forest and expansive droughts might be the feral response that Mother Nature is giving to the big tech industry compensatory projects.
From somewhere to elsewhere: data infrastructure and socio-environmental conflicts
Since its inception, the figuration of ‘the cloud’ has often been portrayed as an immaterial entity existing ‘somewhere’ beyond spatial and geographic constraints (Google, 2006). Our perception of the digital tends to embody a frictionless, distant and boundless spatial bias, especially when viewed from the front-end—that is, through the two-dimensional screens of our devices. In this paper, we align with the definition by Edwards, Copper and Hogan (2024), who describe data centers ‘not necessarily as a place or object, but more as a relationship—or rather, a constantly shifting assemblage of relationships, objectified in particular ways, in particular places’ (p. 433). If the definition of what constitutes a data center is not fixed and should instead be understood as a situated assemblage, this challenges us to develop new theoretical frameworks capable of resisting dominant imaginaries that portray data centers as placeless, neutral, and immaterial hubs of information.
This shift in research perspective emphasizes the impacts of the materiality of digital data (Crawford and Joler, 2018; Cugurullo et al., 2025; Taffel, 2023; ; ) adopting a back-end view that investigates what happens behind the screens. Media critical scholar Mél Hogan (2015, 2018) argues that understanding these structures within their context has become a relevant element of critical infrastructure studies. Hogan's work largely examines the theoretical interactions between the environment and data infrastructure (Hogan and Vonderau, 2019). This comprehensive approach has shed light on the various social, political, and environmental challenges triggered by the presence of data centers, revealing complexities that span across several dimensions and disciplines, as other scholars have shown (Brodie, 2020a, 2020b; Holt and Vonderau, 2015; Lehuedé, 2022a; Munn, 2023; Velkova and Plantin, 2023).
This body of literature shares a common feature: most of the authors and covered cases come from the so-called ‘Global North’. There are a few precedents of research conducted from and for the ‘Global South’, particularly in Latin America, which this paper aims to contribute to. The data center as a subject of study for the region remains somewhat arcane despite the presence of hyperscale data centers in Chile and Mexico dating back to 2015. Sebastián Lehuedé (2022a, 2024), also a media scholar, has written about the growth of data infrastructure in Chile, portraying it as an extractive activity. His scholarship brought the attention of independent journalist Paz Peña (2023), who has discussed the cases of Google data centers in Quilicura and Cerrillos. Both projects gained attention due to local communities’ mobilization such as Movimiento Socioambiental Comunitario por el Agua y el Territorio and Resistencia Socioambiental Quilicura (MOSACAT and RSQ hereinafter). Both groups coin the term ‘socio-environmental’ as these territories are currently sites of social and environmental conflicts due to the resource extraction caused by data center operation. Mexico and Brazil have also become relevant case studies illustrating how struggles emerging from data infrastructure operations are mushrooming across Latin America (see Valdivia, 2024; Lima, 2025). This article contributes to media and critical infrastructure studies (Brodie, 2020a, 2020b; Parker, 2019; Rone, 2023) by shedding light on the environmental compensatory projects and initiatives associated with data centers. In doing so, our paper aims to enrich and further expand the growing field of data center studies.
Despite compensatory projects by data centers being an area rarely examined in depth, there are some exceptions in the critical data center literature. Patrick Brodie (2020b) analyzed Apple's failed data center project in Ireland to reveal how compensation strategies operate in corporate infrastructure development. Brodie showed that the local community cared more about the expected economic benefits than environmental concerns, hoping for indirect compensation rather than relying on clear plans. From a compensation perspective, Apple's approach was both muted and vague, offering symbolic gestures like donating iPads to schools, while rumored training programs were never formalized in official documents. Rather than establishing clear compensation mechanisms, Apple sought to position the data center itself as the main benefit to the community, exemplifying what Brodie describes as a corporate strategy of providing ‘visibility and hope’ (2020b, p. 12) in place of concrete compensatory plans. His study is especially valuable here as it exemplifies how compensatory strategies unfold within the Global North context. However, such strategies do not manifest uniformly in the so-called Global South.
Mexico and Chile are emerging as key players to become global digital hubs driven by three factors. First, both governments are incentivizing data center investments through strategic plans, legislation, and economic initiatives. At the same time of writing of this article, Chilean president Gabriel Boric announced the prospect of 28 possible data centers projects in the country (Ministerio de Ciencia, Tecnología, Conocimiento e Innovación, 2024). Simultaneously, the Governor of Queretaro is also announcing 26 data infrastructure projects to be developed in the future (Estrella, 2024). Second, its geographical position provides a benefit for digital infrastructure. Queretaro functions as a vital internet backbone linking Mexico City to the United States, drawing significant hyperscalers and cloud providers like Amazon Web Services and Microsoft. At the same time, Chile provides connection through the Humboldt Cable, which will establish Chile as a digital gateway to the Pacific by connecting to Sydney in 2026. Third, both countries are underscoring the environmental impacts of data centers. Although these announcements often include promises of environmental and social sustainability, environmental activists and local communities in Mexico are pushing back against these projects. One of the reasons is that data centers in Queretaro are extracting water from aquifers, yet 8 out of the 12 aquifers in the region are already overexploited and at risk of depletion (Conde, 2025). In Chile, Google pledged to offset its carbon emissions associated with the expansion of its data center through a voluntary compensatory reforestation while overlooking the ongoing water impact of its operation and disregarding local communities’ desire to protect their drying wetland.
While both cases are situated in the Global South with bodies of water under pressure, they differ in the way compensatories strategies are designed and unravel in the territories. A ‘feral’ approach can help us see beyond the binary of Global North versus Global South, insisting instead on the specificities of each context where data infrastructures, territories, and communities are entangled in uneven and unpredictable ways. By focusing on how social and environmental compensation projects unfold, a thirsty forest and drought expanses, we can gain deeper insight into the intricate relationships among big tech operations, state interests, local communities, and territories, particularly within the promises of environmental and social responsibility.
Ferality and patchy approaches: data infrastructures in the Anthropocene
As defined by the Oxford Dictionary of English, feral means an animal that is ‘wild, untamed, brutal, savage’ or a plant that is ‘uncultivated’. Feral refers to forms of beings that escape the domestication of life to the wild. This term has historically carried pejorative connotations, often used to negatively label species that transgress human and civilizational norms. For example, a pig that has not been domesticated and lives in the wild is typically referred to as a ‘feral pig’, in contrast to other domesticated pig species. However, the meaning of ferality has recently shifted, particularly among those attuned to contemporary environmental consciousness (Krause, 1993). This transition is deeply connected to the concept of the Anthropocene, a foundational framework in political ecology and science and technology studies, which seeks to explain the planetary-scale impacts of human activity (Moore, 2016). While the Anthropocene highlights the profound impacts of capitalist extraction, dispossession and accumulation on natural landscapes (Harvey, 2012), the idea of the feral has emerged to rearticulate how certain species and environments resist and escape human control.
Within political ecology, ferality is used as a socio-political concept to account for the relation between wildlife and the environment, accounting ‘for dynamic trajectories and processes that escape human domestication’ (Beauté and Dehaut, 2023, p. 1509). In the current climate crisis, nature has become even more out of human control and humanity is now harvesting the unpredictable consequences of its capitalistic adventures. Hurricanes are becoming more unforeseen and brutal; flash floods are untamed in the Mediterranean causing serious devastation and ruination in cities near the seashores; droughts are extending their prevalence with epochs of no rain causing struggles for communities and farmers. However, to the best of our knowledge, these dynamics remain largely underexplored in the context of data centers. While the digital industry celebrates the rapid development of technologies such as generative AI, the escalating climate crisis and its feral effects compel us to critically question the usefulness and sustainability of such large-scale infrastructure projects today. In this context, we turn our attention to the feral effects of human infrastructure, specifically, the compensatory projects now supporting generative AI and related technologies. As data centers become central to digital societies, by accumulating capital, commodifying nature, and reinforcing extractive logics, this paper investigates the feral effects observed of environmental and social compensatory strategies implemented in these local contexts.
Anthropologists Bubandt and Tsing drew upon the concept of ferality to explore how nature resists and reclaims an abandoned coal mine in Denmark, contributing to scholarship on the Anthropocene. Their study demonstrates how nature reappropriates the site, regaining territory once governed by capitalist activity. In their work, Bubandt and Tsing (2018: 1) propose to extend the concept of ferality beyond animals and plants, offering a new perspective on feral dynamics, which they define as ‘cascading effects of more-than-human negotiations’. Building on this framework, the Feral Atlas (2024) expands the project beyond Denmark, exploring feral effects with human infrastructure across the globe. They revisit the concept of ferality: a territorial symptom by capitalist infrastructure and its operation. The feral emerges after years of human dominion over nature and evolving beyond human oversight. For instance, an otter can become feral when it chews through the cables that connect a town in the south of France to the Internet; carbon dioxide could become feral when its accumulation in the atmosphere is large and uncontrolled; an innocuous virus could become feral when its mutation pushes the whole planetary population to a lockdown. We contribute to this framework on ‘ferality’ by investigating the feral effects posed by data centers in Latin America.
Centering our analysis on feral effects leads us to propose a shift in focus from global phenomena to localized analysis. While we acknowledge that the Anthropocene and, more concretely, data center expansions have global and planetary impacts, the study of feral effects advocates for local and patchy frameworks (Tsing, Mathews and Bubandt, 2019). In our specific context, we argue that global and planetary approaches to data infrastructure fail to reveal social and ecological entanglements; instead, they tend to homogenize them. Patchy approaches, proposed as a concept in landscape ecology, counter the overemphasis on the global, homogeneous, planetary nature of the Anthropocene (Tsing et al., 2024). This means identifying what is happening in particular places to distinguish heterogeneous effects through local observations that contribute to the global object of study, e.g., the Anthropocene or the climate crisis. Thus, a patchy approach argues for situated and embodied encounters, where each data infrastructure constructs its own feral patch: the particular ways that each assemblage unfolds.
Within the specific literature on data centers, previous literature has focused on global and planetary perspectives, which are essential for understanding transregional dynamics and the broader reach of data centers. In this vein, Lehuedé (2022b) coined the concept of ‘planetary data turbulence’ identifying environmental concerns as one of three phenomena generating such a turbulence. Drawing on patchy approaches, we argue that such frameworks may not fully account for how these concerns are heterogeneous across different territories and cannot account for the dramatic effects in specific territories. For example, regarding heterogeneity, data centers might exacerbate water struggles in drought-prone areas, while contributing to biodiversity loss elsewhere due to acoustic pollution. In a similar vein, Tironi et al. (2024) ‘planetary AI’ approach operates within abstractions that flatten the local into global calculations. While data centers currently account for 1.5% of the world's electricity consumption in 2024, in some territories, such as in Ireland, this consumption rises to 20%-22% of the electricity grid (D’Ambrosio et al., 2025; Central Statistics Office, 2025). Revealing AI's environmental impacts through planetary perspectives or generalized global metrics erases the specific stories of bodies of water, such as the aquifers of Quilicura and Queretaro, each with its own hydrosocial cycle (Linton & Budds, 2014). In the words of Bubandt and Tsing: ‘abstract calculations […] are not enough; we need to know how the actions and reactions of living and non-living things change the landscape’ (2018, p. 1). These actions, we argue, are accounted for through the patchy approach.
But how are the concepts of feral effects, data infrastructure, and patchy approaches connected? And, what can their intersection reveal about the uneven and contingent ways data centers and their compensatory projects shape local environments? In this article, we contend that data centers, as human infrastructures, are projects that transform and reconfigure natural and non-human components provoking feral effects. However, the feral effects within the data center context become visible to the anthropological eye only when we move beyond global perspectives and approach them through patchy analyses (Tsing et al., 2024). Moreover, within this shift toward patchy approaches, compensatory projects become contested sites of analysis because they rarely operate as neutral or technical fixes. A feral lens helps us analyze how any compensation project can drift away from their intended purpose. Thus, the shift we propose in this paper is from data centers to their environmental and social compensations intended to mitigate their impacts. The feral responds to this definition by enacting situated assemblages that confront the compensatory logics of data infrastructure. These feral patches offer the potential to inspire more situated and environmentally conscious forms of analysis, grounded in the environmental challenges faced by local communities and their territories.
From planetary to local data centers: unveiling the feral consequences of data infrastructure in Quilicura (Chile) and Queretaro (Mexico)
Chile and Mexico are becoming two relevant data centers hubs in Latin America, but at the same time they are also experiencing severe feral effects due to the climate crisis. While Quilicura (Chile) and Queretaro (Mexico) are becoming key data center hubs, both regions are facing severe droughts, putting their aquifers at risk of depletion (Garreaud et al., 2020, González-Sosa et al., 2013). This has raised controversy due to the large amount of water that data centers require for their cooling systems. According to the Environmental and Energy Study Agency, a large-scale data center consumes more than 18 million liters of water per day (Yañez-Barnedo, 2025). Other estimates suggest that a data center requiring one megawatt of energy can consume approximately 72,000 liters of water per day, implying that a fifty megawatt infrastructure could use over 3 million liters of fresh water daily. 1 As a response, data centers are announcing mitigation strategies to mitigate environmental harms and bring benefits to local communities.
Our patchy approach aims to investigate the feral effects of their compensatory projects in these two sites, focusing on their emerging feral effects. Methodologically, we combine thick description (Leeds-Hurwitz, 2015) with mixed methods (semi-structured interview and fieldwork visit). We analyzed 27 policy documents, environmental reports, and media coverage related to the two case studies, and conducted 32 semi-structured interviews with relevant stakeholders in Chile and Mexico during 2023 and 2024. We initially invited five participants, whom we identified as key actors within our context. To do so, we examined reports, journalistic articles, and official documents related to data infrastructure in both selected locations. We then used snowball sampling to identify additional contributors for interviews. The interviewees included data center directors, local policymakers, and individuals from communities living near data infrastructure sites in both countries. Moreover, we also combined insights from interviews with fieldwork expeditions that help us to observe the territory and carefully analyze its feral effects (Tsing et al., 2024).
Enmeshed wetlands and thirsty forests in quilicura (Chile)
In September 2012 Google announced the construction of the first data center, PARAM 1, in Latin America. Chile was the chosen country and Quilicura the exact location. Historically, Quilicura is the largest industrial zone in Chile and, at the same time, home to the most important urban wetlands in Santiago metropolitan region. Before deploying their infrastructure, Google ran geographical analysis and implemented tools to better understand the water bodies within the territory such as the World Resources Institute's Aqueduct Water Risk Atlas and the World Wildlife Fund's Water Risk Filter. However, as they explain in their 2023 Water Risk Framework Whitepaper (Townsend, 2023), these tools did not reflect any specific detail about the water source or any local context to inform more precisely about how their data center could impact Quilicura aquifers. At first sight, Quilicura seemed the perfect site to host a data center: a booming and stable economy with an apparent rich wetland territory.
In Quilicura, Google/Data Luna 2 was not enforced to submit an ‘environmental impact declaration’ (DIA). Chile's Environmental Impact Evaluation Service (SEIA) approved the project through a ‘relevance consultation’ mechanism, an optional procedure conducted prior to formally entering the SEIA. This process determines whether, based on the information provided, a project must submit a DIA. In this case, Google's Data Luna request was given the green light and no environmental evaluation was needed to begin the construction. PARAM 1 started functioning in 2015. In September 2018, Google announced its expansion, PARAM 2 and in 2020, a new data center was proposed in Cerrillos municipality.
This time the environmental regulation in Chile required an assessment for both: PARAM 2 and Cerrillos data center. As a result of the Chilean social outburst in 2019, local communities strengthen their coordination to protect their territory. One of these coordination efforts involved monitoring every environmental impact report publicly available within the SEIA website. Through this process, MOSACAT, a local community based in Cerrillos, gained access to information about the water consumption of Google's data center. Aware that the proposed use of 169 liters per second was excessive given the region's ongoing water crisis, they challenged the project and successfully pressured the company to change its planned cooling system (Lehuedé, 2024).
Despite the access of intended consumption through the environmental assessment documents, no one knows how much water a built data center officially consumes on a daily basis in Chile. The only available estimate about data center water consumption is found in an article published by Rodrigo Vallejos (2022), a young law student member of RSQ that compiled every data center environmental report submitted on the SEIA. In PARAM 2 assessment file, Google has the right to extract 50 liters of water per second; this means 1.576 million liters per year, 3 representing 12.6% of the water consumption of Quilicura communities. Moreover, in 2010, the Santiago Metropolitan Region was officially declared a ‘mega drought’ zone (Garreaud et al., 2020). Within this region, Quilicura has been identified as a hotspot for heatwave activity, recording temperatures as high as 39.7°C, the highest in the past decade (Corporación Ciudades, 2025). What Google believed to be a water-rich environment has proven to be a mirage: a wetland ecosystem now under severe pressure from climate change, further aggravated by data centers consuming the wetland every second. To this day PARAM 1 & 2 is still operating and working as promised, a data center that ‘serves users of the region in a better way’ in a ‘high-tech building that maintains the availability and continuity of services 365 days a year, 24 h a day’ (Data Luna, 2012). Ten years have passed and for RSQ, the arrival of big tech companies has not reflected any benefit or improvement in Quilicura, leaving the neighbors with a passive and distant building complex that for them represents the evaporation of the wetland phreatic water ecosystem.
Known to very few, for PARAM 2 environmental assessment evaluation, Google/Data Luna decided to collaborate with local authorities and invest in voluntary environmental compensation projects. In 2019, Google Quilicura municipality inaugurated the ‘Quilicura Urban Forest’ project. As found on Google's Quilicura webpage the project was situated in a ‘former indigenous site that fell into neglect and disrepair’, and continues, ‘the Urban Forest was brought back to life by Google and local Chilean employees who planted trees and invested time helping with upkeep and landscaping to create a beautiful urban oasis’ (Google, 2025). According to Chile's legal framework all voluntary environmental commitments have to declare the location of the project, a milestone to verify a proper compensation and precise indicators. Google/Data Luna's initially proposed the following voluntary environmental compensations. First, archaeological monitoring during excavation. Second, to train its personnel in cases of an emergency. Third, to notify any fuel spills and/or leaks from back-ups generators (Data Luna, 2017). These proposals were rejected as they don’t represent proper environmental commitments but rather regulatory compliance and obligations that Google/Data Luna had to carry out. Eventually an agreement was reached with a donation of local wildlife. The location would be discussed with Quilicura's municipality during PARAM 2 construction phase (Data Luna, 2017). No community participatory events were held to receive feedback on Quilicura's local environmental struggles or desired compensatory projects.
At its inauguration day, Juan Carrasco, mayor of Quilicura declared that ‘the dream comes true: today we have a new green lung in Quilicura. There are five hectares that will become a space where neighbors, together with their families, will be able to enjoy nature and its benefits, improving their quality of life’ (Cambio21, 2019). Quilicura Urban Forest was presented as a project aimed at mitigating the carbon footprint of PARAM 2 building construction. By planting 1500 native species, an infographic sign states that a 150% offset of the carbon emissions generated by the construction was achieved, normalizing PARAM 1 & 2 water consumption. Legally they are not obliged to provide proof of how this 150% carbon offset took place. The only milestone agreed upon was the species donations and the precise indicator was a count of every plant donated. Species such as Quillay, Maitén, Peumo, Huingán, Romerillo, and Espino were planted at an early stage of growth and therefore require significant supervision, care, investment, and paradoxically water.
To this day, the Quilicura Urban Forest remains an abandoned project, and legally neither the municipality nor the multinational is taking care of the project. We inquired about who holds responsibility for the state of the Urban Forest. No entity assumed accountability: Google fulfilled its voluntary commitments, leaving the matter in the hands of the municipality. However, each local administration operates with its own agenda and budget priorities. Quilicura, a low-income area with limited municipal resources, tends to prioritize more pressing issues such as housing and public safety, leaving the management of the Urban Forest as a very low priority (Escobedo et al., 2005).
After two fieldwork visits to the project in January 2023 and 2024, we could observe no urban forest has flourished in Quilicura. A feral situation emerged where we encountered an abandoned closed lot marked by sparse vegetation and ground blanketed in yellowing grass took over. Near the entrance there is a sign that reads ‘Welcome to Quilicura Urban Forest’ which includes a map and details about the project (Figures 1 and 2). A sign surrounded by scattered shrubs and dry branches that witness a state of struggle or complete abandonment at first sight. For the design of the urban forest, Google contacted a non-profit organization who received funds the first year for seed planting and irrigation system. To this day, the future forest has been left in the hands of a local organization that must seek funding year after year to replace water tanks, hoses, and sprinkler: equipment that is frequently stolen as water stress in Quilicura continues to worsen. Without the sustained efforts of this group, the urban forest would likely have completely withered by now. Unironically, one of those funds is Google Data Center Community Impact Program. In 2021 they won with the project ‘Quilicura Urban Forest 2.0’ proposing new species and plantations to bring native pollinators (Entnerd, 2022). Whoever a thirsty forest persists, accountable for the feral reaction of a soil disconnected from its aquifers.
Entrance to the Quilicura Urban Forest, January 2023. Source: Author’s photograph, taken during fieldwork.
Sign reading ‘Welcome to the Quilicura Urban Forest' inside the Urban Forest, January 2023. Source: Author’s photograph, taken during fieldwork.
When you enter Google's Quilicura data center webpage, you encounter the company's own Impact Study for Quilicura data centers (Google, 2024b). There, Quilicura Urban Forest continues to be highlighted as a key achievement. When we interviewed locals about the urban forest, many responded with confusion, most had never heard of it, nor imagined that Quilicura even had a forest. This is not surprising, as the site is hidden behind industrial warehouses, far from residential neighborhoods, making it highly unlikely for anyone to pass by or engage with it. There is no sense of community ownership or connection to the project, as it was never rooted in the desires or needs of local residents. Meanwhile, local groups such as RSQ and Corporacion Ngen continue to work tirelessly to protect and restore the local wetland. This dissonance between corporate abstract communication and territorial reality illustrates how compensatory projects derails. This contrast highlights the need for accountability mechanisms that not only enforce transparency regarding the environmental impacts of data infrastructure but also strengthen participatory processes to determine whether such compensatory projects will truly benefit the territory. This means moving beyond symbolic mitigation actions to address the situated needs of local communities and their territories.
Drought expansions and water crisis: the data center boom in Queretaro (Mexico)
In November 2023, rural and Indigenous communities protested in front of the council of Queretaro, in Mexico. The local communities from Maconi and Escolasticas claimed that water reserves have been depleted due to intensive activity of large-scale industrial parks, where data centers are also operating. In the last two years, the Governor of Queretaro has invited more data infrastructure projects to be deployed in this region under the paradigm of progress and technological innovation. At the time of writing, Microsoft is set to deploy three data centers in Queretaro to meet the growing demand for services such as ChatGPT. 4 The Governor of the State of Queretaro, together with the United States (US) Embassy in Mexico, has announced what is being described as the largest investment ever made in Latin America by a US company: over US$5 billions for the development of a new data center hub in Queretaro by Amazon. 5
The 10 data centers that are already operating in Queretaro are currently demanding 160 MW, and the 18 new projects will increase this electricity demand to 899 MW (Valdivia, 2024), implying that water consumption could rise to more than 70 million liters of fresh water a day. Despite this estimation, there is a lack of public information about how much water data centers are using; from where they are extracting water; and which local communities could be impacted. In contrast, this state is among the Mexican regions most severely affected by drought. According to the local grassroot organization Bajo Tierra Museo del Agua 6 : ‘In 2022, Queretaro was the only state in Mexico with 100% of the surface area affected by drought, with 16 of 18 municipalities in severe drought’ (Rico-Moran et al. 2023).
One of the main claims of local communities in Queretaro is the right to access water (see Figure 3). While journalistic investigations have suggested that data centers are extracting water from aquifers (Al Jazeera, 2025), the government of Queretaro has developed alternative infrastructure to source water from other areas, including the mountains of Maconi. This alternative project consists of a 123-kilometre pipeline designed to pump water from Maconi to the industrial parks of Queretaro. As part of the project, authorities also promised to improve public infrastructure in the Maconi region by providing residents with access to potable water in their homes, a resource they had historically lacked. However, seventeen years later, this promise remains unfulfilled. Although the pipeline, now regarded as one of the most significant hydraulic infrastructure projects in Mexico due to its length, is completed, the communities of Maconi and Escolasticas still do not have access to piped water in their homes. It is this long-standing neglect that has led to ongoing local protest during November 2023.
The community of Maconi and Escolasticas protesting in front of the city hall of Queretaro during November 2023. The poster reads in Spanish: ‘It is not drought, it is plunder’. Source: Own during fieldwork.
Aware about the increasing bad reputation of data centers guzzling water to communities, this industry is organizing strategies to reinforce their environmental and social responsibility. One such strategy was the project titled ‘Planning for a better quality of life in data center communities’, in partnership between UN Habitat and Microsoft. This initiative, framed within the 11th Sustainable Development Goal (SDG), sustainable cities and communities, aimed to improve the living conditions of local communities situated around data centers in Queretaro. As stated in their website, this partnership sought to: [D]evelop participatory planning processes in the territories where new data centers will be built, in order to develop recommendations and an action plan to make communities more prosperous, sustainable, inclusive and resilient, and to integrate them into the dynamics of economic, urban and social development. UN Habitat, 2022
From an environmental perspective, the report notes that the region ‘has been affected by droughts and water shortages in recent years’, with farmers and low-income households among the most impacted (UN-Habitat and Microsoft, 2022a: 45). In fact, the spatial analysis found that in some communities, only between 0% and 2.8% of households had access to piped water services, indicating that a significant portion of the population lacks water access in their homes (UN-Habitat and Microsoft, 2022a: 82). The report ultimately concluded that severe drought constitutes a serious environmental risk that could increase the vulnerability of communities living near data centres (UN-Habitat and Microsoft, 2022a: 107).
The second stage of the methodology involved co-designing actions based on the conclusions drawn from the spatial analysis, with the aim of addressing the needs identified by local communities. To this end, UN-Habitat and Microsoft organised two online meetings (held under COVID-19 restrictions) and one in-person session. Participants were invited from local government, academia, the private sector, and civil society. Notably, however, Bajo Tierra Museo del Agua, the local grassroots organisation that has consistently investigated water-related struggles in the region, was not included in the process. A further critique is that the gender balance among participants was markedly uneven, with women representing only 22% of those involved.
In the report documenting this stage, participants identified potential risks for the community, including ‘lack of access to safe drinking water due to overexploitation of water sources’ and ‘overexploitation of natural resources’ (UN-Habitat and Microsoft, 2022b: 26). In response, one key action proposed during the sessions was the reuse of water as a means to ‘guarantee the human right to safe drinking water’ (UN-Habitat and Microsoft, 2022b: 33).
The participatory methodology resulted in 21 interventions across eight communities. These interventions were categorised into three main areas: public space, mobility, and outdoor/indoor facilities, with a total budget of US$3.9 million (UN-Habitat and Microsoft, 2022c: 48). Projects included the construction of sports centres, street repairs and lighting, and the enhancement of communal infrastructure. For example, one intervention involved refurbishing a public basketball court at a cost of US$250,000. The most expensive intervention (10th intervention in Table 1) involved road paving.
Summary of urban interventions proposed by microsoft and UN habitat, totaling USD 3 million (2021c).
However, none of the interventions addressed the region's pressing water-related challenges, despite earlier reports (UN-Habitat and Microsoft, 2022a; 2022b) identifying drought and water scarcity as critical environmental risks. While the initial spatial analysis and co-design sessions clearly highlighted access to water as a major concern, the final set of interventions focused solely on recreational and infrastructural improvements, such as sports and communal facilities, and roadworks. This omission raises questions about the responsiveness and priorities of the project in addressing the most urgent needs of the affected communities.
During our fieldwork, conducted in October 2024, we visited all 21 proposed intervention sites and conducted interviews with residents from the municipalities of El Marques and Colon, specifically in the communities of La Loma, Agua Azul, Coyotillos, Viborillas, Galeras, La Esperanza, and Urecho. We identified the exact locations where interventions were reportedly carried out and observed that the condition of the streets and buildings remained unchanged and, in some cases, had further deteriorated (see Figure 4). In our interviews, participants reported that they had not witnessed any interventions by UN-Habitat or Microsoft in their communities, despite the projects being listed in the official report. When asked about the participatory workshops referenced in the documents, most claimed they had not been invited to take part. Moreover, when asked about their access to water and electricity, several participants reported experiencing more frequent blackouts and water access issues over the past two years, roughly coinciding with the time when data centers began operating in Queretaro.
Communal house in Urecho used by the community to organise meetings and events (Colon, Queretaro). The first image (a) is a screenshot taken from the UN Habitat-Microsoft report (UN Habitat and Microsoft 2022c, p. 24) and the second image (b) is a photo taken during our fieldwork in October 2024.
Based on our fieldwork conducted during November 2023 and October 2024, we conclude that communities in Queretaro are experiencing significant water struggles, which are being exacerbated by both climate change and resource-intensive industrial activity in the region. However, none of the 21 interventions proposed by UN-Habitat and Microsoft addressed any issue related to water scarcity. Given the substantial water demands of data centers, alongside the long-standing water struggles reported by local communities such as Colon, El Marques, Maconi and Escolasticas, it is noteworthy that any intervention aimed at mitigating water scarcity in the region. Notably, few participants in the participatory workshops themselves identified the overexploitation of water sources as a key risk in future scenarios.
The feral effects of data centers: thirsty forests and expansive droughts
Feral effects, as Tsing et al. argued, ‘are visible only when infrastructures are studied in their materiality’ (2024: 39). Patchy approaches towards the materiality of data infrastructure reveal how water is the element that triggers feral effects. The two case studies presented in this paper show how ‘challenges of more-than-human livability arise’ (Bubandt and Tsing, 2018: 6). Despite initial promises of economic and technological progress, made by both local authorities and the data centre industry, the transformation of these regions into data centre hubs is producing feral effects, introducing dramatic and often disruptive transformations to local ecosystems and nonhuman life.
In Chile, Google's ‘Quilicura Urban Forest’, a project designed as a voluntary environmental commitment tied to PARAM 2 construction carbon emission, made visible the ferality of a wetland: a thirsty forest withering. This initiative has faced criticism from local communities due to inadequate maintenance, a lack of meaningful involvement, and the neglect of their pressing concern: water replenishment for the wetland. Communities who are working hard in preventing Quilicura's wetland drying process are mobilizing water into the wetland. Quilicura feral patch is an entanglement about how three elements: PARAM's 24/7 cooling towers, a thirsty forest and a drying wetland ecosystem, apart from each other, are connected through underground waters (see Figure 5). In contrast, Queretaro feral patch is an on-going expansive droughts that is likely to intensify, potentially mirroring the pace at which the data centre industry continues to grow in the region (see Figure 6). Moreover, local authorities have deflected responsibility by placing it on CONAGUA, the national agency responsible for managing water resources and issuing water-use licences (Al Jazeera, 2025). Queretaro is currently considered the driest state in Mexico, and is simultaneously celebrated for hosting one of the largest aqueduct infrastructure projects in Latin America. However, this infrastructure reproduces unsustainable water management models in rapidly growing urban areas (Valdovinos & Romero, 2025), thereby facilitating the continued expansion of data centres by the big tech industry and further intensifying pressure on local water bodies.
Quilicura feral patch of data infrastructure, made by the authors.
Queretaro feral patch of data infrastructure, made by the authors.
Our observations suggest that while the data centre industry demonstrates a willingness to publicly announce their mitigation strategies and invest in various initiatives, these projects fall short in two key ways. Firstly, they fail to publicly acknowledge the material complexities and specific socio-environmental conditions of each territory. Secondly, they lack clear mechanisms of liability and accountability for their environmental impacts. The ways in which these territories are being transformed remain insufficiently analysed and recognised, thereby failing to address the root causes of environmental and social responsibility.
Adopting a patchy approach to the data infrastructure context requires acknowledging and carefully studying the relationships between these infrastructures and ‘the beings that respond to them’ (Tsing et al., 2024: 11). Following our fieldwork, this patchy approach led us to identify one such nonhuman entity that warrants closer attention: the aquifers. According to journalistic investigations (Al Jazeera, 2025; Association Press, 2024), data centres in both locations are drawing water from underground sources. This raises concerns about how aquifers, as nonhuman entities, are already responding, and will continue to respond, with feral effects to the intensifying demands of present and future data infrastructure. In both Quilicura and Queretaro, we observed signs of aquifer depletion. Unlike rivers, whose environmental impacts are visible to the human eye, aquifers remain largely hidden, making their degradation more difficult to detect. We contend that future works on the feral effects of data centers should focus on the visible but also invisible components of local patches, either human or non-human.
Beyond mother nature accountability: a feral patch
The burgeoning number of hyperscale data centres constructed in Latin America over the past five years has not only raised growing awareness of their environmental impacts, but has also prompted mounting critiques. Latin America, a region historically harmed by colonialism and capital accumulation through dispossession (Dussel, 1995; Romero, 2021), has also witnessed a large number of communities defending their territory (Alvarez-Villareal, 2024). Within this scenario, the data centre industry has promoted mitigation strategies in response to these critiques. This paper has analyzed two compensatory projects within the contemporary growth of data centres in this region: the ‘Quilicura Urban Forest’ and the ‘Planning for a better quality of life in data centre communities’.
The first conclusion we arrived at is that the data centre industry is designing compensatory projects in Chile and Mexico that are neither delivering promised benefits nor considering situated knowledge (Haraway, 1991). Both the project in Quilicura and the one in Queretaro depend on the financial resources of big tech companies and international organizations, which influence the projects’ objectives and survival more than they promote meaningful engagement with territorial specificities and community needs. Together, these cases expose the limitations of mitigation strategies that neglect the needs of affected territories and communities. This observation, gathered during our fieldwork, resonates with Alvarez-Villareal and Coolsaet's (2020) critique of Western environmental justice perspectives applied in the Global South. According to them, the environment is ‘often defined through Western ways of thinking’ (2020: 50). In both cases, big tech companies reproduce Western approaches to mitigate the environmental and social impacts of their infrastructure, without acknowledging local knowledge or lived experiences.
Second, we realise the importance of remaining attentive to local and patchy contexts while maintaining an awareness of the global and planetary dimensions of data infrastructure. We argue that generalised perspectives on data infrastructure risk obscuring the specific impacts this industry imposes on particular regions. Although both Quilicura and Queretaro have emerged as significant data centre hubs in Latin America, each facing water-related challenges, a locally grounded and nuanced approach reveals the heterogeneous nature of these effects. Although Quilicura requires data centers to publicly disclose environmental impact assessments under, Queretaro has no such regulation. Yet in both locations, local and national authorities lack the regulatory mechanisms to ensure that environmental assessments are tied to enforceable legal obligations.
Third, ferality invites us to critically consider local strategies that genuinely engage with the impacts of data centers beyond their compensatory measures. Our observations lead us to conclude that a patchy approach is particularly effective in the study of data centers, as it reveals the uneven, situated, and often overlooked consequences of technological expansion. While global frameworks identify broad environmental impacts, they obscure the specific ways that each region unfolds to infrastructural interventions. Learning from Quilicura and Queretaro, it becomes evident that compensation projects cannot be understood as generic reforestation projects or cement plazas.
To conclude, this paper proposes a patchy approach, in the sense of recognizing the fragility of each territory within the climate crisis. The techno-solutionist scenario promoted by ‘Mother Nature’ commercials must be challenged and contested. What proves crucial is to remain attentive to local communities as knowledge holders and not merely passive subjects (Alvarez & Coolsaet, 2020). Each testimony matters, a lived experience in the territory, capable of reimagining how a data center or its compensation should be implemented in a specific territory. This approach allows us to remain politically and locally conscious of data centers and their forms of compensation, engaging with human and nonhuman beings impacted by these infrastructural projects worldwide in grounded ways.
Hightlights
Big tech companies’ environmental initiatives emphasize global sustainability criteria rather than addressing local environmental concerns in the areas where their infrastructure is located. Data centers in Latin America are inflicting significant detrimental effects on the environment including water scarcity while their compensatory initiatives frequently fail to adequately tackle local environmental challenges. The paper presents the concept of ‘feral effects’: unforeseen ecological repercussions arising from the interaction between data center infrastructure and compensation initiatives with local environments highlighting the shortcomings of corporate environmental strategy. The paper advocates transitioning from a global ‘net zero’ strategy to a ‘patchy feral approach’ that takes into account specific local socio-environmental settings when developing data center compensation initiatives in areas affected by climate change.
Footnotes
Acknowledgements
This research was supported by the British Academy (SRG2223_231263), the Chilean National Fund for Scientific and Technological Development (Fondecyt) Grant No. 1210006 and the Chilean Millennium Science Initiative Program Grant No. NCS2022_065.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Ethical approval and informed consent statements
Ana Valdivia has received ethical approval by the Social Sciences and Humanities Interdivisional Research Ethics Committee (IDREC) at the University of Oxford (Reference: OII_C1A_23_060).
Nicolas Diaz Bejarano has received ethical approval by Research Ethics and Safety Unit at Pontificia Universidad Católica de Chile (Reference: ID 241126002).
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Fondo Nacional de Desarrollo Científico y Tecnológico, Chile Millennium Science Initiative Program, British Academy, (grant number 1210006, NCS2022_065, SRG2223_231263).
