Simulation Study of Future-Obligation-Aware Workload Admission for Coupled Compute-Thermal Infrastructure

This simulation study will evaluate and attempt to falsify a future-obligation-aware workload admission and supervisory control framework for high-density AI computing infrastructure. The framework treats accepted compute workloads as creating physical resource obligations that may persist after workload execution, particularly unresolved thermal obligations associated with heat storage, heat rejection, and useful heat recovery. The study will model candidate workloads and propagate their predicted electrical, thermal, hydraulic, water, storage, heat-rejection, heat-recovery, reliability, and service-level consequences over a future control horizon. The simulation will test whether incorporating unresolved downstream physical obligations into workload-admission decisions improves the controller’s ability to maintain feasible closure pathways under system constraints. A central hypothesis is that thermal storage moves a thermal obligation through time rather than eliminating it. Stored heat therefore remains an unresolved obligation until useful recovery or heat rejection is measured or otherwise verified. Candidate workloads may be admitted, deferred, limited, relocated, or rejected according to whether their resulting obligations can be closed through feasible downstream pathways. The study is designed for falsification rather than confirmation. Results that show the obligation state provides no meaningful control information beyond conventional state variables, produces no material improvement in constraint compliance or feasible workload admission, or cannot be implemented without unacceptable computational or physical-model uncertainty will weigh against the proposed mechanism. This study follows the Version 1.0 technical disclosure published by Aleeya Kim, Human Layer AI, DOI: 10.5281/zenodo.22818983. The framework remains an unvalidated research hypothesis and this registration does not claim demonstrated performance, engineering readiness, patentability, or freedom to operate.