Chiller plants are among the largest energy consumers in semiconductor fabs, yet they are often operated conservatively to ensure reliability. This presentation looks at how data-driven optimization and digital twins can help balance energy efficiency with operational constraints such as redundancy, maintenance schedules, and variable production loads. Using a real-world case study, it shows how AI-enabled, closed-loop control can support more stable operation, reduce energy intensity, and give facilities teams clearer insight into how their utility systems are performing in real time.