Abstract
Abstract
Circular-economy research has centered on materials, carbon and energy, while water circularity remains fragmented and rarely linked to data-center water demand. The gap matters across the Sustainable Development Goals: the build-out pursued under SDG 9 claims directly on SDG 6, 12 and 13. This paper develops a stock-coupled water–material flow analysis (WFA–MFA) framework integrating ISO 14,046 water-footprint assessment with Monte Carlo-based Pareto screening. Cooling-water demand is coupled to the installed hardware stock and refresh cycle; embodied (scope-3) fabrication water is included; and AWARE scarcity weighting is applied at each component’s own basin. For a 100 MW reference facility, 10,000 draws propagate uncertainty in cooling and power effectiveness, embodied intensity, fab recycling and grid carbon. Both indices increase with performance. Against an evaporative baseline, counting operational cooling electricity only, sealed closed-loop cooling reaches high water circularity but incurs a carbon penalty that leaves it dominated, within the sampled space, in 78% of draws by direct-to-chip and 53% by immersion. Embodied water constrains low-evaporation designs, supplying about 55% of the water they mobilize, a share robust across the disclosed ranges. Operational water circularity in a stressed basin is assessed an order of magnitude below an identical facility in an abundant one. All figures are illustrative, conditional on the assumed parameter ranges and screened from a finite sample. The boundary omits cooling-hardware and dielectric-fluid manufacture, which favors the liquid architectures; no comparative superiority is claimed. Water nonetheless belongs on the same circular accounting footing as carbon in efficiency mandates and siting rules.