Trane and Eaton pair power and cooling in one AI data center design

Trane Technologies and Eaton have bundled their cooling and electrical equipment into a single reference design for AI data centers, claiming a 15 percent gain in energy efficiency and a cut in copper use of up to 80 percent. Data Center Dynamics reports the companies also say installation costs less than with conventional low-voltage design.

The design is part of Trane's Continuum Rubin DSX and Eaton's Beam Rubin DSX platforms, both built for compatibility with Nvidia's DSX architecture for AI facilities. Trane's chief technology and sustainability officer, Mauro J. Atalla, says the pairing delivers a coordinated design that accelerates deployment and lets customers plan to scale. Eaton's Michael Regelski, senior vice president and chief technology officer for the electrical sector, describes the work as moving from reference designs to unified systems teams can deploy repeatedly, integrating Eaton's medium-voltage power systems and white space thermal management with Trane's thermal management architecture.

The companies also say the design aligns with Nvidia's Omniverse DSX blueprint, a direct route into projects already built around Nvidia's specifications. That alignment is where the offering gains traction. Regelski's language points to the market: 'AI-factory deployment at scale' is the sales pitch — data centers built like manufactured products, not one-off engineering projects.

Two claims deserve scrutiny before the industry treats them as settled. The 15 percent efficiency figure is the ambitious one; a verified gain that size would move operating costs. The 80 percent copper reduction, if proven, would change the bill of materials for electrical distribution. The report presents both as vendor claims, without third-party test results.

For investors funding AI data center construction, reference designs are schedule tools. They compress the time between breaking ground and drawing revenue by removing engineering iterations from the critical path. The trade-off is concentration: a single architecture ties the power-and-cooling package to Nvidia's DSX roadmap. If that blueprint changes, the Trane-Eaton package must follow.

Both companies say they will extend the design as liquid cooling and direct-current architectures become mainstream. That commitment concedes the system is arriving in the middle of a technology transition. The next version will be adapted to the next standard, and the two headline numbers will get revisited before the industry settles on a default AI data center. The first facility actually wired and chilled to this reference design will put the 15 percent and the 80 percent to the test.