Rising compute demands, higher power density, and more complex cooling and energy strategies are reshaping AI data center design and driving the need for integrated system-level engineering. 

In this second webinar of a series, a deeper dive into the energy infrastructure surrounding AI factories is discussed. In particular, we will show how a modular, fast-running, multi-fidelity, multi-physics model of a microgrid can be created in GT-SUITE. We will explore how fungible designs with various behind-the-meter energy sources can be created virtually and analyzed to allow energy systems to be scaled and swapped, under a wide variety of locations and operating conditions. Lastly, we will discuss how economic factors can be added to the analysis, allowing for a comprehensive approach to long-term facilities planning.  

Next-generation AI workloads introduce highly volatile, "spiky" load profiles, where an entire cluster can swing between 30% and 150% load in milliseconds, exposing data center power chains to unprecedented risks, including massive inrush currents, severe transient voltage droop, and campus-wide grid flicker.  

In this next installment of our series, we transition from macro-system planning to the deep electrical and thermal infrastructure of the AI data center. 

In this session, we demonstrate how GT-SUITE enables a unified chip-to-chiller workflow, allowing engineers to model component behavior in high fidelity, optimize cooling system performance within the context of the complete data center, and evaluate how component-level decisions influence overall system operation.

We will also showcase how integrated electrical, thermal, and controls models can be used to study cross-domain interactions, fine-tune control strategies, and minimize power and water consumption while maintaining reliable performance.