Trane and Eaton launch joint AI data centre reference design
Trane Technologies and Eaton have announced a joint reference design that integrates thermal management and electrical systems architecture for high-density AI data centres. The two NYSE-listed companies say the unified approach can reduce installation costs by up to 30%, cut copper use by as much as 80%, and deliver combined energy efficiency gains of up to 15% compared with conventional low-voltage designs.
The reference design is built around medium-voltage power distribution and is aligned with the NVIDIA Omniverse DSX Blueprint for AI data centres. It is included in both the Trane Continuum Rubin DSX and the Eaton Beam Rubin DSX platforms, with Eaton supplying power distribution for the Trane thermal management architecture. The companies say the coordinated approach replaces what they describe as a traditionally siloed, manual design process with a single, pre-validated system that data centre teams can deploy repeatedly.
The design
Mauro Atalla, Senior Vice President and Chief Technology and Sustainability Officer at Trane Technologies, said the collaboration was intended to give customers the ability to "accelerate deployment, improve efficiency and confidently plan to scale for the future." Michael Regelski, Senior Vice President and Chief Technology Officer of Eaton's Electrical Sector, said the design advances reference specifications into unified systems "teams can deploy repeatedly."
Vladimir Troy, Vice President of AI Infrastructure at NVIDIA, said the integrated approach "supports the robust, scalable foundations enterprises need to unlock the full potential of generative and reasoning AI." The press release cites a forecast that global data centre capacity may nearly triple by 2030, with AI driving approximately 70% of that growth. Neither company disclosed an independent source for that projection.
The design is described as forward-compatible with emerging liquid cooling technologies and direct current architectures, though no specific product roadmap or certification timeline was provided. No named customer deployments or signed contracts were disclosed.
Market context
The power and cooling challenge in AI data centres has become one of the defining infrastructure constraints of the current build-out cycle. Rack densities for GPU clusters running large language model workloads routinely exceed 100 kW, with some next-generation configurations targeting 200 kW or beyond. That shift has made medium-voltage distribution and liquid cooling essential engineering decisions rather than optional upgrades.
The Trane-Eaton collaboration enters a market where hyperscalers are developing proprietary power and cooling architectures in-house, while colocation and enterprise operators rely on third-party suppliers. Schneider Electric and Vertiv are the most prominent existing players offering integrated power-plus-cooling reference architectures, and both have moved to align products with NVIDIA's reference designs. The alignment with NVIDIA's DSX Blueprint is a commercially significant choice: NVIDIA has positioned its DSX ecosystem as a de facto integration standard for AI factory builds, and vendor certification against it carries practical procurement weight.
On the sustainability side, a reported 80% reduction in copper use is a materials-cost and supply-chain argument as much as an environmental one. Copper procurement has become a meaningful cost variable in large data centre builds, and medium-voltage designs reduce the cable cross-section required at distribution level.
Neither Trane nor Eaton has disclosed the power ratings of the reference design configurations, pricing, or the names of any pilot customers. The efficiency and cost figures cited are vendor-stated maximums relative to a conventional low-voltage baseline and have not been verified by an independent third party.