DIGITAL INFRASTRUCTURE is the spine of modern innovation – from the clouds that fuel e-commerce and artificial intelligence (AI), to the everyday apps. Yet, beneath the gleaming facades of hyperscale server parks lies a fast-unfolding crisis. Data centres have become the new “gold” of investment, attracting a global rush of tech giants and governments alike. But this large-scale infrastructure planning and spread carries high, underexamined costs across several fronts. Currently, cloud platforms run much of the internet, bringing each terawatt-hour of data processing from the working of coal plants, concrete cooling towers, and plunged aquifers.
As nations and corporations race to build more data centres, are we locking in ecological and social harms that our laws and institutions are ill-prepared to manage?
A global data centre rush
Tech titans and states view data centres as high-value strategic assets. The industry is already worth roughly $240 billion globally in 2024, and is projected to more than double by 2032. Hyperscale data centres are campus-scale facilities that often span hundreds of thousands of square feet, and dominate global digital infrastructure. Big tech firms like Amazon, Google, Microsoft and Meta have driven rapid expansion in hyperscale data centre capacity to meet the rising demand for cloud and AI services. Such facilities have not only served as passive repositories of data but have also become critical digital infrastructure underpinning national competitiveness and security.
Northern Virginia’s ‘Data Center Alley,’ the world’s largest concentration of such facilities, is an example of this shift – processing a substantial share of global internet traffic and meeting the needs of millions of users worldwide. Governments across jurisdictions are increasingly recognising data centres as strategic infrastructure rather than ordinary commercial real estate.
As nations and corporations race to build more data centres, are we locking in ecological and social harms that our laws and institutions are ill-prepared to manage?
Singapore, Saudi Arabia, India and the UAE now bid aggressively to host data centres, offering tax breaks, quick permits, and dedicated power zones. Many Indian states have data-centre policies which offer capital subsidies, power rebates, and single-window clearances to attract investors. In July 2024, an executive at Colliers reported an 80% increase in planned capacity in South India, driven by government incentives and digital demand.
This wild expansion is driven in large part by AI. Training a single advanced model can burn through millions of kilowatt-hours (for example, GPT-4 reportedly consumed over 1.7 million kWh). Projects like Amazon’s Project Rainier and Meta’s $10 billion AI hub in Louisiana attest to AI’s hunger for compute. The demand for powerful, intelligent and sustainable data centres is set to accelerate as AI grows.
Hidden resource footprints
As the rhetoric of digital growth continues, it covers stark environmental realities. Presently, data centres require an estimated 1-1.5% of global electricity to keep themselves functioning, and this percentage is climbing. Researchers warn that by 2030, data centre demand could reach 9% of worldwide electricity use. In one analysis of an illustrative German cloud facility, energy use was projected to rise 20% and carbon emissions 13% between 2020 and 2030. Cornell researchers found that, in the US, AI-driven growth alone could add 24-44 million tonnes of CO₂ per year by 2030 - roughly equivalent to 5-10 million additional cars on the road. Even with aggressive renewables, the study finds that without structural changes, “the AI industry’s net-zero targets [will be] out of reach.”
Data centres require a constant supply of electricity and water. Regions rich in solar and wind energy usually lack water, while cooler regions might rely on coal or nuclear power. Prominent cloud companies pledged to reduce their carbon footprints and become water-positive by 2030. Data centres in hotter climates face a trade-off: use less water at the expense of higher power consumption, or vice versa. Historically, companies have treated water as cheaper than electricity, but the modern demand and supply are shifting this calculus. The International Energy Agency (IEA) cautions that the prominent US cloud companies may miss their own carbon and water goals if the current rate of expansion continues.
We need transparency, but there exists a data deficit. Operational energy use is rarely disclosed in detail. Without open data on usage and emissions, sustainable design is hampered and also hinders the development of strategic solutions for carbon reduction. Tracking and reporting remain minimal. Many hyperscale operators even cite trade secrets to withhold water and energy data. If we cannot even measure the true energy and carbon costs of data centres, how can law and policy hope to manage them?
A digital thirst
If energy is one challenge, water is another. Modern data centres guzzle water for cooling, both directly and indirectly. In liquid-cooled systems, vast quantities of water are circulated to remove heat; even “dry” air-cooling designs usually require significant occasional water use. The upshot is that data centres in some countries, like the US, now rank among the top 10 industrial users of freshwater. In a single year, a 1 MW data centre can consume 26 million litres of water, enough for dozens of families. Extrapolate that to today’s largest sites (100+ MW), and the figures are staggering. A 30 MW centre can draw roughly 780 million litres annually. For context, the Yotta Data Centre Park in Greater Noida, India (30 MW IT load today) could eventually draw over 4.2 billion litres a year when fully built, comparable to the annual consumption of a medium-sized town (85,000 people).
These quantities matter because many data hubs are sprouting in already water-stressed regions. In the US, new AI data centres have disproportionately appeared in drought-prone states. Bloomberg News reports that 160 new AI data centres opened in the US over 3 years in areas with high water stress, a 70% increase from the prior period.
In a single year, a 1 MW data centre can consume 26 million litres of water, enough for dozens of families. Extrapolate that to today’s largest sites (100+ MW), and the figures are staggering. A 30 MW centre can draw roughly 780 million litres annually.
Bengaluru’s ~115 MW of data-centre capacity already uses about 8 million litres of water per day; at the same time, the city just endured its “worst water crisis in 500 years”. In rural Uttar Pradesh, villagers now worry that a new data park will gulp up groundwater and deprive farms and homes.
What is worrying is that Indian states are engaging with the idea heavily. Karnataka, Telangana and Tamil Nadu have rolled out dedicated data centre policies. On 20 January 2026, on the sidelines of the World Economic Forum in Davos, a delegation from Madhya Pradesh met with Google’s Asia-Pacific leadership to explore the possibility of data-centre investments in the state. Yet this eagerness belies a dangerous oversight. India is water-poor, especially in booming industrial corridors.
Data centres often travel where power is cheap, but water is scarce, shifting the burden onto local communities. Bloomberg’s analysis finds that China and India have even larger shares of new data sites in arid zones than the US. In the Middle East, Saudi Arabia and the UAE are actively marketing their sunshine for cloud power, despite record extreme drought. Similarly, in Latin America, Chile and Uruguay have licensed new centres even during the worst drought in their history. In Europe, projects in Spain and Ireland raise concerns, and in the Netherlands, locals have fought Meta’s planned facilities over groundwater fears.
And this gets worse as much of a data centre’s water “use” is indirect. A IEA report notes 60% of data-centre water consumption comes from the power plants that fuel them. One analysis found nearly half of US data-centre servers are powered by power plants in water-stressed regions.
Across the globe, protests and bans are rising as citizens seek accountability. In Uruguay, which is under its worst drought in decades, the government’s decision to allow Google’s new data centres attracted outrage and led to mass protests. In a similar tolling situation, in Oregon, citizen groups filed suit to disclose Google’s water-use metrics. The Chilean government was forced to revoke Google’s permits in water-poor Santiago suburbs temporarily amid public outcry. While these instances are small victories, yet they remain fragmented.
Companies and data centre firms claim to use treated or non-potable water, and some claim to use closed-loop cooling to minimise withdrawals. But these vocal affirmations are hard to qualify as reality and account for ‘on paper’. They disclose little to no information about their electricity and water use metrics. Such secrecy highlights a regulatory cavity in environmental impact assessments and clearances. The absence of institutional checks means that India’s data center wave rides on a tide of silent groundwater draw and hidden energy use. Public discussion of water or grid effects is notably absent.
A geopolitics of infrastructure
On the world stage, the data centre boom is taking on geopolitical dimensions. China and the US are locked in an AI arms race, and infrastructure, one among the five layers in AI, is part of the competition.
China enjoys a power edge. In 2024, it generated over 10,000 TWh of electricity, twice the US output, and is projected to add renewables (on course for 40% of its supply by 2030). Further, industrial power in China costs roughly 30% less than in the US, which makes the country’s kilowatts both cheap and plentiful. Also, the Chinese state builds grids and renewables with speed, due to one strong complementing reason of being unconstrained by democratic dissent. By comparison, US demand for data centre power has historically flat-lined, and new capacity is added only grudgingly. Analysts warn of a looming US shortfall with Morgan Stanley estimating a 44 GW gap in meeting demand for new centres by 2028.
Yet the picture is nuanced. The US has led in chip innovation, and American cloud providers still outspend Chinese firms on AI infrastructure. Reports indicate that Chinese companies may invest $147 billion in AI capital expenditures by 2027 - that is less than Amazon alone. Despite its power advantage, China has lagged the United States in building new data centres, partly because of bottlenecks in transmission –the vaunted “Eastern data, Western compute” strategy has left many idle sites.
These dynamics play out globally. China is actively exporting its model with State-funded energy projects in the Middle East and Southeast Asia. These are paired with Alibaba and Huawei cloudbuilds and are meant to shape a global AI ecosystem friendly to Beijing. The US’s counter is its leverage in chip prowess. For example, securing deals to supply American AI chips and cloud services for Middle Eastern gigawatt projects like the Abu Dhabi Stargate data centre. In this multipolar scene, third countries might often mix with some planning Chinese-built solar farms to power American-made AI servers, or vice versa.
Today’s chosen options on the counts of transparency, siting, and acceptable trade-offs will echo for decades, with consequences, good or bad, to be dealt with by future generations and will have impacts, some of which might be irreparable.
Digital sovereignty fights are entwined with climate risk. The least-stressed constant in this race is what the environmental stakes are. China’s expansion of clean energy is impressive, but its solar gains outpace grid capacity, which causes widespread curtailment. The US also needs new renewables fast. Importantly, without strategic planning, both countries risk overbuilding stranded assets, or worse, burning more fossil fuels to meet the bounds of demand and supply. The US-China rivalry is not simply about chips and servers, but about whose infrastructure planning will push the planetary carbon and water balances.
Local harms and governance gaps
Behind all the promised digital infrastructure, data narrative and predictions lie ignored, unnegotiated communities with concerns. The trade-off health and social cost of data centres is now surfacing. In Texas and Arizona, where drought and grid stress already collide, massive AI server farms can exacerbate issues of local water supply. A Harvard-affiliated analysis warns that particulate pollution from diesel generators and power plants backing data hubs could impose up to $20 billion in health costs per year by 2028 in the US.
Data centres may be quiet, but their upstream footprint can blow in the wind and seep into aquifers. Data centres drive digital inclusion, but they risk silencing un-capitalised political voices. Affected communities often have little to no recourse. Environmental Impact Assessments for data parks are notoriously cursory in many jurisdictions.
Water and air are public commons with both protected under the rights to life, the environment, or the ‘public trust’ doctrines. Whose dignity and livelihood count if a village loses its water resource for a cloud facility? Courts are supposed to mediate trade-offs offered by legislators who deprioritised environmental costs in lieu of economic seduction. Are constitutional values of liberty, equality, and due process being honoured when invisible harms are being imposed on the less powerful?
Are we prepared?
Data centre expansion will sky-rocket, straining natural resources. The growth narrative glosses over critical unknowns. Can we meet climate goals at this scale? Can grids keep up without brownouts or coal-burning? Will water tables collapse in rural India? While we lack authoritative answers, some propose smart siting in windy regions. Others call for mandated reporting or energy taxes. But debates focus on sustaining data centres, not whether they're needed. Is this build-out justified by social utility or corporate ambition?
Data infrastructure intersects with environmental law, human rights, and economic policy. Data centres test how laws respond to invisible costs. So far, that response has been reactive or absent. Can constitutional frameworks accommodate data sovereignty and resource stewardship? What duties do companies owe stressed communities? Where companies outpace regulators, how should assessments evolve? Answers will differ by context.
In this age of massive data hubs, it is only a minute start. Today’s chosen options on the counts of transparency, siting, and acceptable trade-offs will echo for decades, with consequences, good or bad, to be dealt with by future generations and will have impacts, some of which might be irreparable. One researcher puts on the AI and data-centre that choices of this decade will “decide whether AI accelerates climate progress or becomes a new environmental burden.” Will these choices be guided by free and open deliberation, or by ill-intent opaque deals and sunk-cost momentum?
The world has yet to carefully and critically grapple with these tough, coherent questions. Data centres have been multiplying quietly and will continue to do so, despite concerns, as a sign of developmental progress and a reflection of rising industry innovation and user demand. Still, the logic of that expansion plan remains unsettled. We should consider: if environmental futures are being written, taking shape into steel and silicon now, should there not be intellectual and public discourse and pace in law-making? In the years to come, technical silver bullets will be far from providing us with solutions. We need to fundamentally rethink how we govern digital growth.