Trane Technologies has introduced a thermal management system reference design tailored for NVIDIA’s Omniverse DSX Blueprint, aimed at supporting gigawatt-scale AI data centers. According to the company, the solution enables rapid deployment, scalability, and optimized performance for advanced NVIDIA AI infrastructure.
The design supports the thermal and power needs of NVIDIA’s GB300 NVL72 systems and is compatible with upcoming Blackwell and Vera Rubin platforms. As rack density grows, the reference design is engineered to scale efficiently while maintaining energy efficiency and resilience.
Trane Technologies stated that the reference design integrates with NVIDIA Omniverse DSX Blueprint for AI Factory Digital Twins, allowing developers to use OpenUSD-based 3D data aggregation for simulation and deployment of AI facilities. This integration is intended to accelerate design timelines and enhance operational flexibility.
“Power and thermal efficiency are now foundational to next-generation AI infrastructure that meets the demands of reasoning and other inference workloads,” said Dion Harris, senior director, HPC, Cloud and Infrastructure Solutions, NVIDIA.
The announcement follows Trane’s recent update to its chiller plant control programming, expanding its thermal management portfolio with solutions such as scalable liquid cooling platforms, air-cooled chillers with larger capacity and higher ambient temperature tolerance, and new Computer Room Air Handler systems designed for AI applications.
