Trane Technologies, a climate technology company, and Eaton, an intelligent power management company, have announced a strategic collaboration combining advanced thermal management and electrical system design aimed at speeding up AI data center construction, improving efficiency, and lowering costs.
The two companies are introducing what they describe as a first-of-its-kind reference design that replaces the traditionally slow, manual, and disconnected process of designing data center power and cooling systems separately with a single, unified, intelligent system. By advancing medium-voltage electrical designs suited to higher-power-density AI facilities, the companies say the combined approach can deliver up to 15% in combined energy efficiency gains, cut copper use, a major cost driver in electrical infrastructure, by as much as 80%, and reduce installation costs by up to 30%, compared to conventional low-voltage designs, all while speeding up deployment and reducing overall system complexity. With global data center capacity potentially tripling by 2030, and AI expected to drive roughly 70% of that growth, the companies frame this reference design as a more resource-efficient way to scale up data center infrastructure to meet that demand.
The new reference design is built into both the Trane Continuum Rubin DSX and Eaton Beam Rubin DSX platforms, developed in alignment with NVIDIA's DSX platform architecture, with Eaton's technology handling power distribution for the Trane system. By combining their respective expertise into a single coordinated reference design, the companies say data center customers can speed up development timelines through pre-coordinated thermal and electrical power systems spanning everything from the electrical grid down to the chip itself. The approach is meant to break down the traditional practice of designing power and cooling systems in separate silos, instead letting those systems share performance data and respond more dynamically to each other's needs, improving overall data center performance.
Mauro J. Atalla, senior vice president and chief technology and sustainability officer at Trane Technologies, said AI and high-performance computing are reshaping what's demanded of data centers, and that customers want solutions capable of keeping pace with those changing needs. He said combining Trane's thermal management expertise with Eaton's power management technology results in a coordinated design that helps customers deploy faster, improve efficiency, and plan confidently for future scaling.
Michael Regelski, senior vice president and chief technology officer for Eaton's Electrical Sector, said the companies are advancing the industry standard for deployment speed by turning reference designs into unified systems that teams can deploy repeatedly. He said that by aligning with NVIDIA's DSX platform and integrating Eaton's medium-voltage power systems and thermal management technology with Trane's system architecture, the collaboration helps speed up large-scale AI data center deployment.
The joint reference design is built to work smoothly with NVIDIA's Omniverse DSX Blueprint for AI data centers, a framework for planning AI infrastructure. Together, the companies say this combined approach establishes a more consistent, predictable way to plan and deliver the electrical, thermal, and digital control infrastructure needed for AI-driven environments, simplifying setup, reducing risk during deployment, and improving overall performance for current technology, while also being designed to adapt as emerging liquid cooling technologies and direct current power architectures become more widely adopted.
Vladimir Troy, vice president of AI infrastructure at NVIDIA, said AI factories require tightly coordinated power, cooling, and compute infrastructure to operate efficiently at scale, and that by aligning with the NVIDIA Omniverse DSX Blueprint, Trane Technologies and Eaton are helping customers reduce complexity and speed up deployment of next-generation AI data centers. He said this integrated approach supports the robust, scalable infrastructure enterprises need to fully take advantage of generative and reasoning AI and turn data into faster, smarter outcomes.
