Optimizing Data Center Energy Efficiency in the Age of AI

Data center computing power demand is estimated to reach 800 TWh by 2026, roughly equivalent to Japan's electricity consumption, mainly due to the increased use of AI. In this context, data center owners, motivated by operating cost and sustainability imperatives, are looking to reduce their energy footprint through novel approaches.

​​​​​​​In this webinar replay, we will present a holistic Virtual Twin approach to tackle this issue. This involves integrating and coupling HVAC and IT systems into a single evolving model, capturing the behavior of the data center during its operation. This model provides essential information in order to predict and optimize the allocation of load within a room, taking into account its thermal behavior to reduce day to day energy consumption. The same model can also be used to observe and plan for future hardware upgrades, ensuring that the thermal envelope of the room, and its resulting energy consumption post upgrade, remain within specifications.

Throughout the presentation of this Virtual Twin, we will touch upon:

• Modeling and simulation of equipment (HVAC, Direct Liquid Cooling (DLC) system  & servers)
• Modeling of the computer room, including the cooling systems and the IT
• Coupling of airflow simulations (CFD) with DLC systems to improve cooling efficiency
• Extension to IT (application load) / Facility coupling for an holistic approach of the analysis of the energy efficiency