Data center operators report water use as a facility-level question, disclosing on-site metrics while omitting the water consumed in generating the electricity they purchase — their Scope 2 water footprint. This paper quantifies that omission at the eGRID-subregion level and traces its consequences for a standardized data center load. Comparing EPA eGRID 2018 against eGRID 2023 on an identical pipeline, I find that generation-weighted national water intensity fell 17.1% over the five years in which data center electricity demand accelerated, but that the decline was uneven: several subregions gaining generation share fell more slowly than the national average, and SERC Virginia/Carolina — host to the largest data center market in the world by inventory — fell only 7.7%, less than half the national pace. Combining eGRID generation data with consumption-basis water coefficients, I estimate a total of 1,350.6 billion gallons of operational water consumption embedded in contiguous-US electricity generation across all end uses in 2023, and show that an identical data center's embedded footprint varies approximately 3.2-fold by grid region under common technology-factor assumptions. I then construct a Water Dependency Index and adjust it by a territorial water-scarcity score built from the World Resources Institute's Aqueduct basin indicators, on the principle that identical consumption imposes different costs depending on where it occurs; the adjusted index ranks SERC Virginia/Carolina first in the nation. No widely adopted disclosure framework requires standardized, location-based reporting of electricity-embedded water. The paper proposes extending location-based Scope 2 carbon accounting, which is structurally analogous, to water.