Invisible Consumption: Data-Centre and Cloud Water Disclosure Gaps in India's Listed IT Sector

This paper examines water disclosure for data centres, AI workloads and purchased cloud among India’s listed IT companies. Across 20 FY2025-26 Business Responsibility and Sustainability Reports (BRSR), none reports a quantified water figure linked to data centres, AI workloads or purchased cloud, while three report an energy-related data-centre metric. Only TCS quantifies total data-centre energy use, and it also reports the power usage effectiveness (PUE) of its data centres to two decimal places (1.58). TCS’s reported PUE shows that at least one company already has the energy data needed for the denominator of Water Usage Effectiveness (WUE); the water figure needed to complete the calculation is not disclosed. The gap is about to get bigger. TCS’s FY2025-26 annual report describes HyperVault, a subsidiary planning up to 1 GW of AI data-centre capacity in India, starting with a 100 MW first phase. The BRSR chapter of the same report doesn’t mention it. If the full 1 GW were IT load running at full use, illustrative scenarios put its on-site water at about 1.1 billion litres a year withdrawn, at Amazon’s withdrawal-based WUE, or 4.8 to 14.3 billion litres consumed, at Li et al.’s consumption-based factors. The water consumed in generating its electricity on India’s coal-heavy grid comes to about 19 to 48 billion litres a year across the capacity and grid scenarios modelled, roughly 7 to 17 times the 2.83 billion litres TCS reports as its total water consumption. These are scenarios for planned capacity, not forecasts, and no BRSR indicator asks for either number. The twenty companies were coded with a fixed protocol and checked against public CDP water scores. For the companies that own no data centres, the absence of data-centre water figures is expected, because their water use sits upstream with cloud providers, which a companion note examines. For the companies that do run data centres, it is a real gap. The pattern fits missing rules better than missing ability, since some of these companies already report the energy side, though a review of public filings can’t show what any company measures internally. India’s 2023 national water neutrality framework defines a general water-use coefficient per unit of output, which could be adapted to data centres as WUE, but it never mentions data centres, servers or cooling. The paper proposes a five-part amendment to BRSR Principle 6, plus a disclosure comparing permitted and actual groundwater use for data centres in water-stressed areas. Working paper, not peer reviewed. Evidence dataset: https://doi.org/10.5281/zenodo.23590593