Daikin Applied unveils 60 MW AI factory cooling design built on NVIDIA's DSX platform

Daikin Applied announced in Minneapolis on 10 November 2026 a 60 MW cooling reference design for AI factories, named D60-W Water-Cooled Chillers with Dry Coolers. The company says the design is based on the NVIDIA DSX AI Factory Platform with NVIDIA Vera Rubin NVL72 systems and covers cooling from the rack out to the open air.

According to Daikin Applied, D60-W handles a 60 MW IT load split across six deployable blocks, which lets a project expand in steps. Liquid cooling removes 92% of the heat and air cooling the other 8%. A high-temperature loop feeds the coolant distribution units, and a separate low-temperature loop supplies the rack containment units that handle air cooling. Air-cooled, liquid-cooled and hybrid computing can run next to each other, the company says.

The plant uses Daikin Applied's oil-free Magnitude WME magnetic-bearing centrifugal chillers, which come in sizes up to 10 MW, together with dry coolers. The company says that in suitable climates the dry coolers handle cooling through most months and the chillers help only on the hottest days. Depending on the weather, the system runs in dry, hybrid or full mechanical mode: in dry mode the chillers' compressors stop, and in hybrid operation waterside economisers cool the returning water first and the chillers carry what is left.

Daikin Applied says that rejecting heat through dry coolers rather than evaporative cooling towers means no make-up water is needed for towers and keeps water usage effectiveness low. It also says the oil-free chillers do away with oil heaters, pumps and routine oil servicing, which cuts standby energy and maintenance, and avoid start-up delays caused by oil when the chillers come back from standby.

For power problems, the company says its RideThrough and RapidRestore chiller technologies deal with short disturbances and with restarting after a shutdown, and that redundancy is built into every stage between rack and dry cooler to keep cooling running through planned maintenance or a failed unit. Its iDCM supervisory controls run the rack containment units, coolant distribution units, chillers and dry coolers together as a single plant, sharing load between machines, evening out their running hours and switching between the three modes, Daikin Applied says.

Daikin Applied says its own coolant distribution unit and a dry cooler designed for data centres are still being developed, and it expects both to be available in the second half of 2027. It offers commissioning, planned maintenance and round-the-clock emergency response for the design.

Yu Nishiwaki, Daikin Applied's president and CEO, says AI factories exist to convert energy into useful output with as little waste as possible, and puts cooling at the heart of that. Karim Shalash, the company's director of data centre solutions, calls an AI factory a system of systems and says a cooling choice made at the rack affects water temperatures and flow, how the plant runs and how heat is rejected.