New Delhi: At a time when the Centre and states are offering tax breaks and power incentives to draw artificial intelligence (AI) investment, the World Bank warns that households sharing grids with data centres could end up footing their electricity bills.
India along with Brazil, Malaysia, Thailand and China are countries where data centre investment is “having immediate impacts on energy demand and prices”, according to the World Development Report (WDR) 2026, ‘The Promise of Artificial Intelligence’.
“Households that share electricity grids with data centers bear a substantial proportion of the costs incurred to serve these data centers,” it notes.
The report was launched last week by the World Bank Group with the IndiaAI Mission.
Power regulators allow electricity distribution companies to set tariffs that cover their costs and earn a return. The report says this can raise bills for other consumers in two ways.
“First, if utility providers sell energy to data centers at concessional prices, they must recover the difference by increasing prices for other customers. Second, consumers may bear the cost of the infrastructure investments required to serve data centers, given that the distribution of such costs among different groups of customers is often not transparent,” it says.
“Individuals living in places with weak consumer protections are particularly vulnerable to inequitable electricity-pricing practices,” the report adds. “Those that profit from the infrastructure for AI may not bear the cost of building and maintaining it.”
India’s data centre capacity is projected to rise from 2.2 GW in 2025 to 12 GW by 2030, according to consultancy Wood Mackenzie. In this year’s Union Budget, the Centre announced a tax holiday until 2047 for eligible foreign cloud companies that use data centres in India.
States are offering their own incentives. For instance, Uttar Pradesh’s data centre policy, approved in July, exempts data centres from electricity duty for 10 years.
Globally, electricity use by data centres grew about 12 percent a year between 2017 and 2024, over four times the growth in all other sectors, the report says. It is projected to more than double by 2030. Running AI models can use 25 times the energy needed to build them, it adds, “indicating that demand for energy stemming from AI will likely soar in coming years”.
Stress on water resources
Data centres worldwide are projected to consume about 1,200 billion litres of water a year by 2030, up from about 560 billion litres.
These centres, the report says, are being located based on “the availability of energy resources and favorable regulations, rather than the availability of abundant water resources”. “The siting of data centers in areas subject to water shortages places additional pressure on already-scarce resources and threatens the supply of water for local communities,” it adds.
The concern has reached state legislatures. In March, Karnataka IT Minister Priyank Kharge told the Assembly that the southern state was reviewing its data centre policy. “Data centres are a necessary evil,” he said. “But data centres are also heavy water and energy guzzlers.”
“Hyperscale data centres won’t suit Bengaluru, which doesn’t have a port and faces shortage of water,” he asserted.
World Bank’s suggestions
The report proposes that governments make grid connections for new data centres conditional on “greater operational flexibility”. Data centres could be required to cut use during peak demand, shift tasks that are not urgent to off-peak hours, or install on-site battery storage.
It also asks governments to weigh where data centres are built. A facility linked to new renewable power can add capacity, it says, while one “added to a grid that is dependent on fossil fuel or already constrained may increase emissions, worsen congestion, and increase the cost of local electricity”.
The World Bank also cautions developing countries against building data centres to pursue “AI sovereignty”. “Large investments in data centers may also impose other societal costs, such as higher electricity prices,” it asserts.
(Edited by Tony Rai)