In Uttar Pradesh’s Tusiana village, a gleaming data centre stands amid crumbling roads and open drains.
Residents fear that the Yotta Data Centre Park, built to power India’s digital future, could deplete their groundwater.
India’s push to expand AI-ready data centres is colliding with mounting water stress across several regions.
Companies and governments reveal little about how much water these energy-hungry facilities actually consume.
Experts warn that without transparency and regulation, India’s digital infrastructure could come at a steep environmental cost.
Two starkly different worlds converge in a small village in Uttar Pradesh. At the entrance to Tusiana village in Greater Noida, a sprawling 20-acre complex looms along a wide, freshly paved road, its perimeter ringed with high walls, barbed wire and police barricades. Inside, a towering sub-station hums with electricity. Trees and gleaming SUVs line the route, creating a veneer of prosperity. Yet, drive a kilometre further and the illusion dissolves: the roads grow rough, the air turns foul, and residents trudge past open drains and sewage-choked streets — a daily reality in a village still deprived of basic infrastructure.
This is Tusiana, home to around 2,000 people in Gautam Buddha Nagar district. In 2022, it became the unlikely stage for a high-profile inauguration, when Uttar Pradesh Chief Minister Yogi Adityanath and then Union Minister of State for Electronics & IT Rajeev Chandrasekhar opened the Yotta Data Centre Park. A Memorandum of Understanding (MoU) worth Rs 39,000 crore ($4.39 billion) was signed between Yotta and the state to develop six data centres over the next five to seven years. The facility is currently only partially operational, with one of its six data centres open.
While Yotta Data Services touts the facility as “North India’s gateway to the digital world”, village residents say it has brought them little benefit. In interview after interview, residents told Down To Earth (DTE) that the centre neither improved local living conditions nor created jobs. “When the construction is complete, they will not need labourers like us,” says Satish Chand, a local worker.
Residents of the village, however, express a deeper anxiety — over the future of water availability.
India’s digital ambitions are expanding at breakneck speed, and data centres are seen as critical infrastructure to sustain the growing adoption of artificial intelligence (AI). Yet their rapid spread poses a serious challenge: these facilities require vast amounts of water to cool servers, and many are being built in regions already grappling with water stress. In such areas, potable water is often diverted for industrial use, leaving economically marginalised residents more vulnerable.
Field visits by DTE to two emerging data centre hubs in Bengaluru and Gautam Buddha Nagar found that while states compete to attract data centre companies with incentives and subsidies, few are assessing the long-term implications for local water security. The companies, in turn, disclose little about their actual consumption, often cloaking figures in vague or incomplete data.
In Tusiana, shopkeeper Rahul Bhatti is one of the few aware of what this might mean. “Our main source of water is groundwater. For now, there is no shortage,” he says. “But when the whole facility becomes operational, it could draw heavily from the aquifers and reduce water available for us.”
Another resident, Satish Chand, admits he barely understands what a data centre does but recalls seeing deep borewells being drilled — “around 200 feet down” — when he worked as a labourer on-site. Yotta Data Services clarified to DTE that no borewell was dug for any construction or operations activity.
If you’ve ever opened an app on your phone, joined an online class or meeting, saved photos, played games with friends, or used artificial intelligence (AI) to draft an email or write code, you have relied on a data centre.
A data centre is a facility that houses IT infrastructure such as servers, computer switches, storage systems, routers for storing, processing, and distributing data and applications. An AI data centre, on the other hand, can perform high-performance computing tasks, allowing quick data processing and analysis. These facilities are powered by multiple central processing units (CPU), graphical processing units (GPU), and memory chips that are a part of a server.
Though data centres have existed since the 1940s, their transformation began in the 1990s with the rise of smaller computers and the internet. Then came the internet, the IT boom, cloud computing, and artificial intelligence, all of which drove up the need for greater computational power and data storage, fuelling the demand for data centres.
Most of the electricity used by servers is converted into heat, which must be continuously removed to prevent slowdowns or catastrophic failures that could disrupt services like online banking, credit card processing or digital payments.
A McKinsey & Company analysis estimates that by 2030, around 70 per cent of all new data centre capacity will cater to AI-heavy workloads. Generative AI — systems like ChatGPT or Google’s Gemini that create text, images or code — is especially water-intensive because of the more powerful GPUs it uses.
Researchers at the University of California found that after generating a simple 100-word email, an AI chatbot’s cooling systems consume between half and three bottles of water (200–1,500 millilitres), depending on local conditions. Training these models requires exponentially more: Meta reportedly used 22 million litres of water to train its LLaMA-3 model.
Cooling is not the only concern. Data centres indirectly consume water through the electricity they draw from power plants, placing them among the top 10 water-consuming commercial sectors in the US, according to a 2021 study in the journal Environmental Research Letters.
Water for cooling typically comes from groundwater or surface water sources, though some facilities use recycled supplies. In 2023, Google reported that nearly 80 per cent of the water used for cooling at its global data centres was potable. Recycled water can ease pressure on local supplies, but it is no panacea.
“Recycled or non-potable water can help reduce stress on drinking water sources — but only if managed carefully,” Shaolei Ren, associate professor of electrical and computer engineering at the University of California, Riverside, told DTE.
“Treatment and infrastructure are needed, and the environmental impact depends on where and how the recycled water is sourced. In some cases, claims can be overstated, so careful scrutiny is important.”
Back at Yotta’s facility, operations are already geared towards AI-intensive workloads. The IT load capacity, which corresponds closely to the amount of computing equipment a data centre can support, currently stands at 30 megawatts (MW). Researchers estimate that a small 1 MW data centre consumes about 26 million litres of water annually. Extrapolated, a 30 MW facility could use roughly 780 million litres of water a year — equivalent to around 219,700 litres every single day.
To put this in perspective, that daily volume could meet the domestic water needs of approximately 1,630 people in urban India, based on the government’s estimate of 135 litres per capita per day.
And this is just the beginning. Once all six data centres at the Yotta Data Centre Park are operational, projected for completion by 2026, the total IT load capacity will rise to 130 MW, according to Yotta’s own website. That would translate to around 3.3 billion litres annually, or 9.26 million litres a day. Such consumption is comparable to the daily water use of an entire town of about 68,000-70,000 residents — roughly the size of Punalur municipality in Kerala, as per the 2011 Census.
These estimates align with those cited in a report from the Government Digital Sustainability Alliance, an alliance established to improve digital sustainability outcomes for the United Kingdom government and its supply chain, which noted that a 100MW facility is capable of consuming around 2.5 billion litres of water annually.
These figures are indicative rather than definitive, and the company claims it uses only 5 lakh litres per month or 60 lakh litres [6 million litres] annually. They nonetheless alarm residents of Tusiana, who say their groundwater levels have been declining for years. “Two decades ago, we used to get water at 20 to 30 feet. Now, we must dig at least 80 feet,” says Chand.
Though the village receives piped water from the government, he explains that it is often unsafe: “Sometimes cracks develop in the pipes that run close to sewage lines, mixing dirty water with clean. We can’t drink that. Groundwater is our only source.”
Water use in data centres depends on a range of factors like location, climate, water availability, size, and technology involved. Yotta’s website states that its facility uses a chilled-water system for cooling. In a region like Noida, where summer temperatures can soar to 48 degrees Celsius, cooling demand is inevitably high compared to a place like Bengaluru. In winter, when temperatures drop to around 3°C, the facility can draw in cool ambient air, reducing electricity consumption. “This helps reduce power consumption during winter,” the company’s website notes.
Government records reveal little to no information on Yotta’s water consumption. In Uttar Pradesh, data centre proposals are assessed by the State-Level Environmental Impact Assessment Authorities (SEIAA), which grant environmental clearance based on recommendations from the State-Level Expert Appraisal Committees (SEAC) — bodies responsible for screening and evaluating projects.
On October 15, 2020, the SEAC recommended environmental clearance for Yotta’s proposal, approving a water use limit of 120 KLD. The clearance required the company to obtain permission from the Central Ground Water Authority (CGWA) — which regulates groundwater extraction — or from another competent authority if surface water was to be used.
That same day, the SEIAA granted clearance, stipulating that if the project fell within a “notified area” — zones identified as over-exploited, critical or stressed in terms of groundwater — then the company must meet its freshwater needs through alternate or legally permitted sources. Greater Noida is indeed one such notified area. Yet, when DTE sought clarification from the district’s groundwater department, officials did not confirm whether Yotta had permission to extract groundwater.
The company claims that it is not using groundwater or municipal drinking water at its Greater Noida Data Center Park. It uses 5 lakh litres of water per month for cooling, which is supplied entirely through GNIDA’s treated, non-potable water network that is separate from the residential supply.
The Greater Noida Industrial Development Authority (GNIDA), which oversees the city’s infrastructure, said it currently supplies 50 KLD of treated water to the company. Yotta has its own 50 KLD sewage treatment plant (STP) for recycle water for further utilisation in horticulture cleaning. “During the starting of the project, GNIDA provided two separate water connections, one for potable water of 5 kilolitres per day (KLD) and another for STP 50 KLD for construction activities,” the company told DTE.
“The company has made an additional request for treated water,” confirmed Vinod Sharma, GNIDA’s executive engineer. Yotta has not published any environmental, social and governance report either.
“Yotta Data Center cooling system is designed on close loop and air-cooled system, which doesn’t require water for daily operations,” the company told DTE. Closed-loop cooling means water is continuously recirculated rather than consumed or discharged, and only minimal top-up. Air-cooled systems, however, are more energy- intensive.
Yotta’s sustainability policy, available on its website, states that it is committed to “protecting the environment by actively reducing our environmental impact through strategic conservation efforts and resource optimisation”. Beyond this broad statement, there are no concrete details on targets or measures on the website. The policy also claims efforts towards electricity, energy and water savings, but again offers no evidence or metrics.
The opacity around Yotta’s water consumption raises a deeper question: How are these water-guzzling industries being approved and monitored in areas already facing scarcity — and who bears the consequences?
This is part one of a four-part series. Parts two, three and four will dive deeper into India’s data centre boom and its environmental cost.
This story was produced with support from Internews’ Earth Journalism Network.