How is data centre water use independently verified?
Verification requires four things to be true at once: the figure is measured on the pipe rather than estimated from an invoice, measurement is continuous rather than periodic, every instrument carries a trust score so an untrustworthy sensor yields a hold rather than a value, and the underlying records are replayable so a third party can recompute the totals and get the same answer.
The fourth condition is the one that separates verification from monitoring. Building management systems and chemistry-vendor platforms already measure some water at most large facilities, but their records are recalibrated, backfilled and interpolated as normal operating practice, and they carry no signature or chain. They exist to run the plant, not to serve as evidence.
A verified record states the water balance for each interval: withdrawn by source, returned and to where, reused inside the fence, consumed, and the water embedded in electricity shown separately rather than folded into a total. Each interval is signed and linked to the one before it, so any alteration is detectable.
The practical test is whether the party asking the question, usually a council, a regulator or a First Nation water office, can obtain the raw records and recompute the published figure without relying on the operator's word.
Sources
Virginia requires data-centre water-use reporting from 1 January 2027
Virginia HB 496Utah H.B. 76 requires reporting from 2027 with penalties up to $10,000 per day
Utah H.B. 76