Process‐informed co‐optimization of liquid air energy storage and data centers under off‐design thermodynamic operation

Abstract The rapid growth of AI data centers (AIDCs) is intensifying stress on power systems and exposing operators to volatile electricity markets. This work develops a process‐informed co‐optimization framework that integrates liquid air energy storage (LAES) with AIDC operation. Unlike black‐box storage models, the proposed formulation captures LAES thermodynamics through charging start‐up dynamics and off‐design discharge behavior, linking process efficiency with dispatch decisions. The model is formulated as a mixed‐integer program with a robust extension for joint participation in energy and reserve markets. Results across CAISO, PJM, and ERCOT show that LAES–AIDC integration reduces operating costs by 9%–50.9% in energy markets and up to 66.1% when reserve services are included. The analysis reveals that optimal operation is governed not only by price signals but also by maintaining LAES near efficient thermodynamic regimes, with AIDC workload flexibility shaping demand to support process performance.