India's AI data centre expansion strains water supplies

India is rapidly building AI‑focused data centres, with capacity already exceeding 1.5 GW and plans to reach 6.5–17 GW by 2030. The high‑density GPU racks consume large amounts of power and generate substantial heat, making liquid cooling increasingly necessary. Because liquid cooling relies on water, the sector could use up to 37.5 billion litres of water a year, adding stress to cities such as Mumbai, Hyderabad, Chennai, Bengaluru and Pune that are already classified as water‑stressed. Major hyperscalers and the Adani Group have pledged more than USD 167 billion for new facilities, but water availability remains a key constraint.

In New Zealand, a Singapore‑based firm, Datagrid, has secured approval for a NZ$3.5 billion, 280 MW AI data centre near Invercargill, slated to open in 2028. The plant would become the country’s second‑largest electricity consumer after the Tiwai Point aluminium smelter, prompting the Green Party to call for a pause on new AI data centres while stricter rules are devised. Officials are debating whether such centres can operate as flexible grid assets, shifting workloads to periods of high renewable generation to reduce peak demand.

Both regions illustrate how the AI data‑centre boom is prompting policymakers to balance economic benefits with the growing pressure on water supplies and electricity grids.