Sustainability-Constrained Workload Orchestration for Sovereign AI Infrastructure: A Joint Compute-Network Optimization Framework

TL;DR

This paper introduces a framework for sustainable AI infrastructure management that jointly optimizes compute and network resources within strict environmental constraints, ensuring operational feasibility and reduced environmental impact.

Contribution

It presents the Feasible Sovereign Operating Region (FSOR) concept and demonstrates joint optimization improves sustainability and signals infrastructure needs.

Findings

Joint optimization reduces environmental impact compared to baseline methods.

Infeasibility signals indicate infrastructure investment or workload reduction needs.

FSOR characterizes sustainable workload levels under physical and regulatory limits.

Abstract

AI infrastructure has transitioned from a software-centric paradigm to a system tightly bound by physical and environmental limits. Energy availability, cooling capacity, and network connectivity now impose hard operational boundaries that cannot be relaxed through software optimization alone. This paper proposes a sustainability-constrained orchestration framework that treats carbon intensity, water usage, and power capacity as strict feasibility constraints rather than tunable penalties, and that jointly optimizes compute placement and optical network routing in a single closed-loop system. We introduce the Feasible Sovereign Operating Region (FSOR) - a conceptual and operational construct that characterizes the set of workloads a given infrastructure can actually sustain under its physical and regulatory endowment. Scenario-based analysis demonstrates that joint optimization yields lower environmental impact relative to baseline formulations. Infeasibility events, rather than being optimizer failures, constitute precise, telemetry-grounded signals that sovereign AI operation requires infrastructure investment or workload reduction.