AI Data Centers as Programmable Civic Heat Nodes: A Modeling Framework for Workload-Shaped Public-Facility Preheating

This preprint proposes a modeling framework for treating AI data centers as programmable civic heat nodes rather than merely electricity-intensive facilities with recoverable waste heat. The central idea is that AI workloads, cooling architecture, and thermal output can be configured so that recovered heat is shaped toward useful preheating demand in nearby public facilities. The framework models data-center heat as a shared preheating layer across a civic portfolio that may include public swimming pools, hospitals, schools, housing loops, and greenhouses. It introduces a preheating offset function and two evaluation metrics: Useful Thermal Match Efficiency (UTME) and Community Thermal Coverage (CTC). Literature-range simulation results suggest that liquid-cooled AI data centers operating at moderate outlet temperatures can provide meaningful public-facility preheating value. The paper defines the novelty boundary, mathematical model, simulation approach, and validation plan required for future empirical testing with measured workload and facility heat-demand data. This is preprint version 1.0.