The Asymmetric Boundary

Companies that take water seriously still stop counting it somewhere, and this note tries to find where. I looked at nine large companies in four sectors (banking, telecoms, consumer goods including brewing, and IT services), based in three countries. I asked two questions. Does the water used to run the cloud computing they buy show up anywhere in their environmental accounts? And do cloud providers appear among the suppliers they work with on water? The carbon from the same purchase gets counted in two different ways. One company asks its cloud providers for their own emissions data. Another uses a spend-based Scope 3 Category 1 method that allocates supplier emissions according to expenditure, so cloud-related carbon may be included without being separately identified. None of the nine discloses a comparable cloud-water figure in the public documents I reviewed. The named, measured supplier water programmes I found in the sample focus largely on agriculture, particularly barley, rice and hops growers and the regions they come from. No company names a technology, cloud or data centre provider among the suppliers it works with on water in its public reporting. One company, Vodafone, has formally determined that water is not material to its own operations or its value chain. In July 2026, Amazon Web Services (AWS) began giving customers annual water-withdrawal figures for their workloads. The next reporting cycle is the first in which a company using AWS could answer the question with a figure from its provider. Reporting frameworks address water in the value chain, and India’s BRSR Core value-chain framework asks for a water figure from major partners on a voluntary basis. None explicitly names purchased computing, and I found no evidence in the company reports reviewed that suppliers are being asked for a water figure attributable to cloud computing. Research note, not peer reviewed. Companion paper, Invisible Consumption: Data-Centre and Cloud Water Disclosure Gaps in India’s Listed IT Sector: https://doi.org/10.5281/zenodo.23594888 Evidence dataset for the companion paper: https://doi.org/10.5281/zenodo.23590593