Daikin Applied launches 60MW cooling blueprint for NVIDIA AI factories
The new reference architecture provides a standardized cooling system for high-density AI hardware, aiming to simplify data center engineering.
The short version
- Supports a 60MW IT load split across six deployable units.
- Uses liquid cooling for 92% of the load and air cooling for the remaining 8%.
- Features oil-free chillers that eliminate standard oil servicing and reduce standby energy.
- Operates without cooling-tower make-up water to support low water usage effectiveness.
- Targeted availability for the company's own CDU and dry cooler is H2 2027.
Why it matters
As AI chips become hotter and more dense, cooling is becoming a primary bottleneck for scaling data centers. This blueprint helps developers move faster by providing a pre-coordinated system of parts rather than forcing them to source and integrate individual components from different vendors.
A blueprint for AI scale
Daikin Applied has unveiled a new 60MW cooling reference architecture designed specifically for AI factories. This blueprint, named D60-W, is tailored for the NVIDIA DSX AI Factory Platform and the NVIDIA Vera Rubin NVL72 hardware. It provides a starting point for developers to design cooling systems from the individual rack level up to the ambient environment.
The company aims to simplify the complex process of coordinating multiple suppliers and teams. Typically, cooling plans require separate coordination for rack requirements, facility equipment, heat rejection systems, and controls. By connecting these elements into a single system, Daikin Applied intends to help customers move more quickly into the project engineering phase.
Yu Nishiwaki, President and CEO of Daikin Applied, noted that cooling infrastructure is central to the challenge of turning energy into output efficiently. The D60-W architecture is intended to give customers a single partner to work with as they scale their AI operations. It addresses the growing factor of how quickly and reliably AI capacity can be brought online.
Hybrid cooling and two loops
The D60-W architecture supports a 60MW IT load split across six deployable units. It employs a hybrid approach where liquid cooling handles 92% of the load while air cooling manages the remaining 8%. This design utilizes two distinct temperature loops to improve the overall efficiency of the plant.
A high-temperature liquid loop serves the coolant distribution units, which are the primary method for cooling the racks. Meanwhile, a dedicated low-temperature loop serves the air-cooling rack containment units. These units, known as S-Series RCUs from DDC Solutions, capture the remaining heat at the rack and move it directly to the low-temperature loop.
The system is designed to allow operators to run air-cooled, liquid-cooled, and hybrid compute side by side. This flexibility is important because a cooling decision at the rack level has significant implications for water temperatures and flow requirements. The architecture provides a framework for engineering teams to manage these dependencies across the entire facility.
Oil-free chillers and water savings
At the heart of the architecture are Daikin Applied’s oil-free Magnitude WME magnetic bearing centrifugal chillers. These units have capacities of up to 10MW and differ from traditional large-scale chillers that rely on oil lubrication. Because they are oil-free, they do not require oil heaters, pumps, or standard oil servicing, which reduces energy use during standby periods.
The oil-free design also removes start-up delays associated with oil management, allowing the cooling to recover quickly to protect GPU availability. The system can switch between three modes based on outdoor conditions: dry cooling, hybrid cooling, and full mechanical cooling. In dry cooling mode, the chiller compressors remain off, while hybrid mode uses waterside economisers to precool return water.
A significant environmental feature of the D60-W is its reliance on dry coolers instead of evaporative cooling towers. This means the system requires no cooling-tower make-up water, which supports low water usage effectiveness. This setup is particularly beneficial for regions where water conservation is a priority for data center operations.
Centralized control and roadmap
Daikin Applied uses its iDCM supervisory layer to coordinate all components, including the RCUs, CDUs, chillers, and dry coolers, as a single plant. This software layer manages load sharing, balances equipment operating hours, and controls the transitions between different cooling modes. It also provides built-in redundancy and integrates with site management systems for predictive maintenance and energy optimization.
The architecture is designed to scale by adding deployable units as demand grows. Daikin Applied provides the commissioning, planned maintenance, and 24/7 emergency response for these systems. This end-to-end service model is intended to reduce the operational burden on data center operators who are scaling rapidly.
While the blueprint is available now, some components are still in development. Daikin Applied’s own CDU and data centre dry cooler are targeted for availability in the second half of 2027. The company has also expanded its portfolio through acquisitions of DDC Solutions and Chilldyne to include leak-resistant liquid cooling and direct-to-chip capabilities.
What to watch
- The rollout of the Daikin Applied CDU and dry cooler in H2 2027.
- How other hardware providers respond to standardized cooling blueprints.
- The adoption of oil-free chiller technology in non-AI data center environments.
Try it
Look up the 'Low Water Usage Effectiveness' (WUE) metric to see how data centers measure water consumption compared to energy use.
Sources and links
First reported by Data Centre Magazine. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.