Daikin unveils D60-W cooling model for NVIDIA DSX AI factories

The D60-W is positioned squarely against NVIDIA's DSX AI Factory Platform, the design framework that takes the Vera Rubin NVL72 supercomputer as its core. By publishing a reference architecture rather than a single product, Daikin Applied Americas is offering developers, designers and operators a standardized starting point that spans from the server rack all the way to heat exhausted outdoors, with cooling equipment and controls integrated as one system.

The design leans heavily on liquid: 92% of its 60MW IT load is liquid-cooled and only 8% air-cooled, split across six units that can be added incrementally. Two liquid loops at different temperatures let the design route higher-temperature coolant to the CDU while a dedicated low-temperature loop serves the air-cooling rack units, which the company says raises overall efficiency.

The central component is Daikin's Magnitude WME, an oil-free magnetic bearing centrifugal chiller capable of up to 10MW. In the D60-W, dry coolers carry most of the cooling load for much of the year where climate allows, with the chiller stepping in during the hottest outdoor periods. Racks are mainly liquid-cooled through the CDU, while remaining heat is captured in-rack by DDC Solutions S-Series rack containment units and sent directly to the low-temperature loop. Each rack also embeds fire suppression and water-loop protection, and the design accommodates mixed air, liquid and hybrid IT equipment.