There is a recent thread that links together San Francisco, Vancouver, Tucson, Santiago, Dublin, London, Johor, Batam and the Indian port city of Visakhapatnam, also known as Vizag. In a catastrophically fragmented geopolitical framework that defines the world these days, the common element is a major blow against AI data centers in these regions. At a time when India is opening the doors to artificial intelligence (AI) data center investments, global sentiment is quite the opposite due to the environmental impact, including local energy and water demands, that underpin this infrastructure.
The root of this reaction can be best understood from statistics. First of all, the question of expansion of electricity and infrastructure is essential. Secondly, concern about clean water. Fengqi Yu, who holds the prestigious Roxanne E. and Michael J. Jack, who is the chair of the professor, told HT that it is also an issue of fairness in sports.
“AI services can be used nationally or globally, but the burden of water and electricity is local. If a data center uses scarce freshwater while nearby communities face rationing or farmers face irrigation stress, that becomes not just a technology problem but a water-justice problem. The solution is to make water a binding design constraint,” he says.
It may be prudent to disclose expected water withdrawals and consumption, cooling technologies, water sources, and drought contingency plans in data center projects.
According to the International Energy Agency (IEA), global electricity consumption by data centers is poised to surpass 1500 TWh (or terawatt-hours) by 2035, up from about 400 TWh in 2020. In the Indian context, government data shows that power demand will reach 2703 TWh by 2031-32, of which data centers will consume 4% of this, or about 107 TWh.
To put something familiar in context, an individual server rack in an AI data center is about the size of a large refrigerator – by 2027, the peak power demand of each could be equivalent to that of 65 homes.
Ashish Banerjee, senior principal analyst at Gartner, tells HT that India needs to build AI data centers only after carefully considering location and cost. “The next phase should be planned around power and water, and not just around land availability. India’s data-centre capacity is already around 1.5 GW and could reach around 6.5 GW by 2030, water usage will be around 150 billion liters annually in 2025 and is expected to more than double by 2030,” he said on grid preparedness, availability of non-potable water and reliable points out three key considerations of renewable-plus-storage sourcing.
Data centers in India are growing in areas that are already facing severe water crisis. Water security is a primary issue, and allocating huge amounts of potable water to cool AI servers is raising public concern.
Concerns about water scarcity, along with deforestation, form the basis of protests by residents of Visakhapatnam. The greatest instability comes from the proposed facility at Adavivaram-Mudasarlova, a site adjacent to the Mudassaralova Reservoir, the primary drinking water source.
The government, in a note issued earlier, stressed regulation and control of groundwater extraction, including for industrial purposes, to be governed by guidelines issued by the Jal Shakti Ministry, which were last revised in 2023.
There is an important difference between traditional data centers and AI data centers. Traditional enterprise server racks draw 15–20 kW of power, while AI and machine learning workloads use dense GPU clusters that can consume 80–150 kW per rack.
Banerjee says, “The threat is inherently local, not national. A data center may seem small in national water accounts, but in a water-scarce district during extreme summer heat it could become a very visible threat to local citizens and farmers.”
Global examples point in this direction.
Slate Asset Management’s $10 billion “Steelport” data center project in Hamilton, Canada faced opposition from residents due to environmental concerns and the use of water from the bay. Microsoft’s Build developer conference in San Francisco this month saw protests over the use of land, water and electricity by the company’s data centers.
In the US alone, an estimated $64 billion worth of data center projects were blocked or delayed between May 2024 and March 2025 – including a $14 billion project in Arizona and a $24.7 billion project in Virginia.
A Chilean environmental court has suspended Google’s $200 million data center project in the Cerrillos district of Santiago following a lawsuit from residents. The government in Johor, Malaysia, banned new Tier 1 and Tier 2 data centers (facilities with lower efficiency standards) in November. UK energy regulator Ofgem is considering ‘mandatory curtailment’ for data centers by reducing electricity use during periods of grid stress.
Last year, Cornell researchers noted the need for collaboration in a study in collaboration with Sweden’s KTH Royal Institute of Technology, Concordia University in Montreal, Canada, and the RFF-CMCC European Institute for Economics and the Environment in Milan.
He said that even as companies like OpenAI and Google are investing more and more money into building AI data centers faster to meet demand, this is a critical moment for coordinated planning between industry, utilities and regulators to avoid local water shortages and higher grid emissions.
For India, teaching a global lesson
For Cornell U, India needs to focus on two key priorities. First, the strategic sites that will define data centers will be located where the grid can support larger loads, where new clean power and storage can be added, and where cooling demands do not worsen local water stress. Secondly, the fact that most of the capacity expected by 2030 has not yet been built means India has a chance to design sustainability from the beginning rather than having to retrofit it later.
He says, “India can certainly pursue AI leadership, but it should not treat data centers as just a digital-infrastructure issue. They are also large, constant users of electricity and water. The key question is not how fast India builds, but where it builds, at what power, and with what cooling systems.”
Banerjee points to an interesting duality in how AI data centers might be structured. “Ironically, high-density AI provides a technological escape hatch as large-scale AI racks exceed the air-cooling limit, forcing operators to transition to closed-loop direct liquid cooling. Instead of evaporating millions of liters of municipal water per day into the atmosphere, these systems recirculate liquids internally,” he explains.
India has the option of banning open-loop potable-water cooling in high-stress areas, while instead mandating treated wastewater, seawater or closed-loop liquid architectures. For example, Microsoft uses closed-loop cooling systems in its Fairwater AI data centers in Atlanta and Wisconsin (US), Norway and the UK to extract hot liquid to cool it before returning it to the GPU.
“Every aspect of the Fairwater AI data center and networking innovations has been optimized to deliver AI computing power with the greatest efficiency and using the fewest resources,” says Alistair Spears, Microsoft’s general manager for Azure infrastructure.
Cushman & Wakefield says in its latest Global Data Center Market Comparison 2026 report that power availability is a key determinant of data-center growth. They identify electricity infrastructure, grid readiness and transmission as key barriers to expansion.
“The global data center sector is moving into a more performance-driven phase of growth, where access to power, infrastructure readiness and delivery capacity are becoming as important as demand,” says Gautam Saraf, executive managing director, Mumbai and New Business, Cushman & Wakefield.
India’s major data center projects announced this month include Reliance and Meta AI’s 168 MW data center in Jamnagar, Google Cloud expansion in Navi Mumbai and CtrlS Datacenter’s 600 MW mega campus in Navi Mumbai and Chennai.
Mark Zuckerberg, Meta’s founder and CEO, said about the Jamnagar project, “This world-class facility in Jamnagar will help us scale our AI infrastructure globally while deepening our long-term investment in India’s economy.”
No major data center projects have been canceled in India so far, but the sector continues to face delays and localized opposition.
“Carbon intensity across the world’s major data center hubs varies by orders of magnitude. Indonesia, India and Hong Kong (SAR) are among the most carbon-intensive grids, with carbon footprints 62%, 51% and 43% higher than the global average, respectively,” the UN University’s Institute for Water, Environment and Health said in its latest report on the unequal distribution of data center facilities.
“Local and national governments must remember that managing electricity and water use more efficiently is essential to protecting electricity affordability, maintaining system reliability and attracting investment,” says Hongpeng Lei, head of the Climate Change Division, the mitigation arm at the United Nations Environment Program (UNEP).
While the US is already writing the response chapter, India is largely in the euphoria phase. But inevitably, the same questions about natural resources will arise.
“I wouldn’t see this as simply hitting the brakes. But India should avoid the build first, control later model. The better approach is a science-based green-light and red-light framework. Projects that demonstrate clean power access, grid readiness, low water intensity, responsible siting, transparent reporting and fair cost allocation should move forward,” you point out.
Strong regulation is necessary to ensure that projects that do not meet these criteria are not approved and reworked.
The future of AI may depend less on algorithms and chips than on whether society has enough sustainable sources to power the cloud, which is becoming a resource-intensive industry.
