AI-driven data centre expansion poses a triple environmental threat—rising emissions, water consumption, and electronic waste—while the ecological toll of e-waste and the loss of critical minerals and rare earths remain underexamined
Image Source: Getty Images
The unrelenting expansion of artificial intelligence has finally begun to attract meaningful scrutiny regarding the environmental impact, albeit unevenly. An International Energy Agency (IEA) report in 2025 found that data centre electricity consumption has grown 12 percent annually since 2017, with nearly 60 percent of that demand currently met by fossil fuels. Water consumption has drawn comparable concern: estimates indicate that globally, data centres use approximately 560 billion litres of water per year, potentially rising to 1.2 trillion litres by 2030. Another peer-reviewed study reported that in 2025, AI-driven data centres consumed more water than the global consumption of bottled water, estimated at 446 billion litres per year.
Beyond driving emissions and water usage, which have been the primary focus of environmental scrutiny, an important aspect of the data centres’ environmental impact has received relatively less attention: the generation of massive amounts of electronic waste (e-waste) and other hardware detritus. Current systems are not equipped to handle e-waste at this scale. Large amounts of e-waste end up in landfills or are incinerated, with direct consequences for air and water quality at disposal sites, rather than being recovered.
Data centres are already a resource-intensive sector; their energy demands, water consumption, and e-waste streams create compounding pressures. In India, all three are particularly concerning.
Data centre expansion, therefore, carries a triple threat to the environment through emissions, water consumption, and waste generation, resulting in costs that cannot be offset by any margin of profit. In the wake of India’s AI Impact Summit, data centre expansion is being treated as an unambiguous national priority. Prime Minister Modi, speaking ahead of the summit in February 2026, declared, “We invite the whole world's data to reside in India.” On its own merit, data sovereignty is undoubtedly an important issue; the Global South generates a disproportionate share of global data, yet the United States (US) alone accounts for 44 percent of global data centre power capacity. This asymmetry has driven legitimate ambitions around data sovereignty and domestic infrastructure investment across the developing world.
Many, however, argue that expanding data centre infrastructure, primarily to support AI development, is not a direct solution to this asymmetry and could create more problems than it solves. The case for domestic data infrastructure is sound. The toll it takes on the environment deserves more attention. Data centres are already a resource-intensive sector; their energy demands, water consumption, and e-waste streams create compounding pressures. In India, all three are particularly concerning.
AI-related hardware has a short commercial shelf life; GPUs and servers are replaced every two to five years, often faster as successive model generations render previous ones commercially obsolete. Estimates project that generative AI alone could generate up to five million metric tonnes of accumulated e-waste by 2030. Countries operating on older hardware due to US export restrictions on advanced chips face a further penalty, generating up to 14 percent more e-waste per unit of computing output.
AI-related hardware has a short commercial shelf life; GPUs and servers are replaced every two to five years, often faster as successive model generations render previous ones commercially obsolete. Estimates project that generative AI alone could generate up to five million metric tonnes of accumulated e-waste by 2030.
For India, this lands atop a pre-existing waste management crisis; the downstream consequences are not theoretical. Research from the Mandoli industrial area in Delhi, published in a peer-reviewed environmental journal, found heavy metal concentrations in groundwater from informal e-waste recycling sites exceeding both Indian Standards and WHO permissible limits for drinking water. Components without resale value are openly burnt or dumped, releasing dioxins, PM10, and heavy-metal particulates into the air of communities, also contending with coal-fired power and over-extracted groundwater. India generated 1.75 lakh metric tonnes of e-waste in FY 2023-24, with formal processing capacity already insufficient to absorb it. A 2026 CPCB assessment found 17 states and union territories with no registered recycling facilities. Waste that the formal system cannot absorb passes to the informal sector, entering soil and water already degraded by the energy and cooling demands of data centres. The ‘triple threat’ does not result in separate crises; its effects are compounding, and the environment bears the ultimate cost.
Policies that discourage early replacement cycles require no new technology. This, however, requires intent that is currently absent. The Union Budget 2026 proposed a tax holiday until 2047 for foreign companies providing cloud services through Indian data centres. This is not a neutral investment signal. It is the removal of fiscal leverage at precisely the moment when the government has the most power to require responsible infrastructure practices. Mandatory EPR compliance under the E-Waste (Management) Rules, 2022, verified recycling partnerships, and regional facility requirements that close the 17-state gap should not be voluntary addenda to investment deals. There should be conditions of such deals, and taxation should not be abolished for an industry that poses these risks without guarantees of compliance.
The AI bubble may not last much longer, but the debris it generates certainly will.
Even closing the recycling gap offers limited recourse. As research has documented, there are structural limits to recycling, particularly in case of e-waste. Electronics are rarely designed with recyclability in mind, and the economics of proper recycling seldom work in its favour. India is inviting the world to build the digital infrastructure of the Global South. However, replicating the extractive model of the countries that built the first generation of this infrastructure, and inheriting their waste alongside their workloads, will not result in data sovereignty. The AI bubble may not last much longer, but the debris it generates certainly will.
Krishna Vohra is a Junior Fellow with the Centre for Economy and Growth at the Observer Research Foundation.
The views expressed above belong to the author(s). ORF research and analyses now available on Telegram! Click here to access our curated content — blogs, longforms and interviews.
Krishna Vohra is a Junior Fellow in Climate and Energy at the Centre for Economy and Growth. His research spans across resource governance, circular economy, ...
Read More +
PREV
