In this decade, we witness an unprecedented expansion of digital infrastructures, including artificial intelligence. However, scholars and practitioners around the globe are consistently raising concerns around how the cloud is not weightless and it has serious implications on land, livelihoods and communities. To understand these concerns, the Digital Empowerment Foundation’s (DEF) research Just AI, Just Land: Socio-Environmental Implications of Data Centers on Communities examined how these developments are unfolding in peri-urban regions, centring on community perspectives. In it, we outlined how expanding data centre infrastructure is influencing land use, water access, and livelihood patterns, while tracking how local communities navigate. This was a far more complex story to document than it appears on the surface, shaped by multiple layers of policy, infrastructure, and lived experience. What we are sharing here is only a snapshot of a much larger and evolving reality.
In the last few decades in India, many governments and institutions have come and gone, policies have been announced, revised, rebranded, and yet some problems have politely refused to leave. Chief among them is the inadequate planning of resources like land, water and power in a way that actually keeps up with reality. The result is a steady stream of not-so-cheerful headlines and reports: 21 Indian cities expected to run out of groundwater by 2030, two-thirds of households still navigating some form of energy poverty, and persistent inequalities across social groups despite decades of reform.
So, naturally, into this already well-stocked situation arrives the newest guest: the artificial intelligence (AI) data centre boom. Large, gleaming infrastructures promising intelligence, speed and the future, while quietly consuming vast quantities of the very basics we were already running short of: water, land and electricity. What makes it more interesting is the geography of it all. These facilities tend to land in places where resources are already stretched, but conveniently cheaper, and where policy incentives make the deal even sweeter, in the Global South.
It is an old story with a new costume; earlier it was labour, now it is land, water and energy doing the heavy lifting. And to take that “heavy lifting” quite literally, a single AI data centre can consume up to 20 lakh litres of water daily, which in a day is equal to the daily needs of 6,500 households. When it comes to power one AI-focused 100MW data centre can consume as much as electricity required for almost 750,000 Indian households. They further require large tracts of land, which is scarce and unevenly distributed due to socio-historical inequalities in ownership and access across communities in India.
The geography of growth
To be fair, the money is real. The “revolution” has done what it promised: it attracted serious investment. As per the reports, by the end of 2025, the sector had already accumulated around USD 126 billion, with projections climbing to nearly USD 180 billion by 2026. In the last few years, the AI data centres started coming up in metropolises like Mumbai, Bengaluru, Hyderabad, Noida, Chennai – again cities that sit in the red zone when it comes to the above resources, especially water. While the corporate numbers and reports term these cities as “kingmakers”, the reality is quite complex.
Many researchers and journalists slowly started asking questions like: what exactly is the pleasure in hosting these facilities here, when parts of the Global North seem far less enthusiastic about having them in their own backyards? They also started documenting what local communities were saying when these facilities were built near their homes. As these conversations grew, researchers and journalists also began paying closer attention to what local communities were actually saying when these facilities came up near their homes.
Following the ground reality
Those narratives led us to a Rest of World report on Mekaguda, Telangana, where residents alleged illegal land encroachment by Microsoft. As an organisation working on digital technologies and AI, we decided to look into this more closely, not just to understand what communities were saying, but why they were saying it.
We focused on two locations in Telangana, Mekaguda and Begarikancha, in the Rangareddy district. The idea was to centre community voices and place them alongside existing research and expert discussions. Over eight months, we read existing policies and literature, spoke with researchers, lawyers, experts on power, water, and land, as well as journalists, academics, community representatives, health workers, teachers and IT professionals. We asked a central question: what socio-environmental impacts could these AI data centres create, and how would they affect communities? Alongside this, we conducted focus group discussions to understand lived experiences, impacts, and whether communities were included in decision-making processes at all.
Experts pointed to several concerns. These infrastructures are likely to strain already limited resources, and even where scarcity is not immediate, the burden will fall more heavily on marginalised communities. The policy environment also raises significant questions. In Telangana, data centres are offered extensive incentives, including exemptions from key environmental and labour regulations, self-certification in place of inspections, and the bypassing of laws such as the Maternity Benefit Act and Minimum Wages Act. By being classified as “essential services”, they are also guaranteed uninterrupted power, subsidised electricity, and cheaper land. Efficient, yes, but it does make one pause and ask, efficient for whom?
There is also very little transparency. It is still not clear how many resources these centres consume, how many people they actually employ, or what their long-term stated contributions to the state will actually be.
What communities are saying
On the ground, communities raised concerns that were both immediate and layered. Land acquisition was a major issue. Some farmers reported filing petitions against Microsoft over alleged encroachment and waste dumping. Others spoke about inadequate compensation, especially for assigned lands, which are government-allocated lands, often given to individuals from historically disadvantaged communities with restrictions on sale. As per community’s opinion, the compensation, in many cases, did not match the actual value of the land. This has had a deeper impact on communities classified as Scheduled Castes, Scheduled Tribes and Other Backwards Classes, which are communities that have historically faced social and economic disadvantage in India. Many of them were the primary holders of assigned lands, making them more vulnerable to displacement.
In terms of water and power, the impacts are uneven but visible. In one village, people are already buying drinking water. In another, water pollution has affected cattle and agriculture. There are also fears around radiation, even if direct impacts on electricity are not yet visible. At the same time, these are not just local resources. Water and electricity are drawn from larger regional systems, which means the stress may not always show up immediately next door, but it is felt somewhere.
There are also aspirations. Many community members believe that such large infrastructures will bring jobs. But in reality, opportunities have been limited, and mostly restricted to low-paying roles like security or maintenance.
The question that remains
One positive change we witnessed was how awareness evolved over time. In the beginning, only a few people in the villages knew what data centres were, many thought they were just another industrial setup. But over the months that changed, and by the final phase, during a community workshop where we shared research findings, people were listening and were asking questions. Important ones.
What now? Now that you have informed us about this, what will you do? What is the solution?
This question did not come only from the communities; during the study, even as we were thinking through recommendations, we were asked similar questions.
Is there such a thing as good or bad AI? Will frameworks and guidelines actually make AI better, more ethical? If yes, how?
We would leave that question with all of you as well. It is not an easy one to answer. Many might say that the infrastructure is already here, so what can really be done. But that does not excuse us from engaging with the harder questions. It does not remove the responsibility of going back to communities with something more than just documentation.
Because the question still stands. What is the solution?
Maitri Singh, Dr. Arpita Kanjilal and Mili Dangwal work at the Research and Communications Division of the Digital Empowerment Foundation.
This piece is a version of information shared by DEF as part of the project "Just AI, Just Land: Socio-Environmental Implications of Data Centers on Communities" for the "Seeding Change" column. This column presents the experiences of APC members and partners who were recipients of funding through APC's subgranting programmes, supported by the Swedish International Development Cooperation Agency (Sida), and of subgrants offered through other APC projects.
Did this story inspire you to plant seeds of change in your community? Share your story with us at: communications@apc.org