Midea Building Technologies is positioning integrated cooling infrastructure as a response to the growing demands of high-density artificial intelligence computing, introducing a power-cooling hyperconverged architecture and a full-stack cooling portfolio. The technologies combine power systems, energy storage, magnetic bearing cooling sources and liquid cooling distribution to address infrastructure requirements across AI data centers. The approach targets operators balancing higher computing densities with energy efficiency, equipment constraints and demanding operating environments. Midea Building Technologies presented the solutions at Asia Data Centre World 2026 in Singapore, placing coordinated power and thermal management at the center of its infrastructure strategy.

Integrated Architecture Connects Power and Cooling

The power-cooling hyperconverged architecture brings together infrastructure components that traditionally operate as separate systems. It integrates power supply, energy storage, magnetic bearing cooling sources and liquid cooling coolant distribution units (CDUs) into a unified infrastructure stack. The design connects these elements from the early stages of data center planning and construction, allowing operators to coordinate electrical supply and thermal management before facilities enter operation. This integration aims to reduce dependence on conventional mechanical cooling through coordinated operation of magnetic bearing cooling sources, free cooling and liquid cooling terminals.

The approach reflects a wider infrastructure challenge for AI operators: cooling performance increasingly depends on how equipment works together rather than on individual component efficiency alone. Coordinating power and thermal systems can help operators manage the demands of high-density computing while pursuing lower power usage effectiveness (PUE). However, actual efficiency gains depend on site conditions, equipment configuration and operating profiles. For facilities facing rising rack densities, the architecture offers a framework for planning power and cooling as connected infrastructure requirements.

Magnetic CDU Targets Smaller Equipment Footprints

Midea’s full-stack cooling solution spans cooling sources, distribution systems and terminal heat dissipation. A central component is its Magnetic CDU, which combines a magnetic bearing cooling source with a distribution unit instead of relying on the conventional separated “chiller + passive CDU” configuration. According to MBT, the integrated system can reduce equipment footprint by up to 70% and support a PUE below 1.2 under suitable operating conditions. The figures highlight the company’s focus on reducing space requirements while improving the efficiency potential of high-density computing environments.

The footprint claim could matter for operators working within constrained buildings or seeking to accommodate additional computing capacity without proportionally expanding supporting infrastructure. Combining cooling generation and distribution also changes how operators approach equipment layouts and system integration. The reported PUE figure, meanwhile, represents a performance target rather than a universal outcome across installations. Site-level performance will depend on climate, workload, cooling configuration and other facility systems.

Air-Cooled Chiller Addresses Water Constraints

MBT also introduced an Air-Cooled Magnetic Centrifugal Chiller designed for hot environments where water availability is limited. The equipment delivers a coefficient of performance (COP) of up to 5.4, according to the company’s specifications. Air-cooled systems can offer an alternative where operators want to reduce dependence on water-based heat rejection, although overall energy performance remains sensitive to ambient temperatures and operating conditions. The product extends MBT’s portfolio toward locations where climate and water constraints complicate conventional cooling strategies.

For large AI clusters, the company also highlighted its Industrial-Grade CDU, which provides 2.6 megawatts of cooling capacity, equivalent to approximately 739 refrigeration tons, at a 3-kelvin temperature difference. The unit targets liquid cooling distribution across large-scale AI clusters and high-density racks. Its stated capacity places the product within the infrastructure segment serving substantial thermal loads. Meanwhile, the specified 3K temperature difference provides an important operating parameter for evaluating the unit’s suitability within a particular liquid cooling design.

Keppel Collaboration Demonstrates Free Cooling Potential

MBT cited its collaboration with Keppel at the Gui’an Midea Cloud Data Center in China as an example of its cooling technologies in operation. The project combines MBT hardware with Keppel’s digital platform to coordinate infrastructure performance. According to the companies’ project details, the facility provides up to 7,654 hours of annual free cooling while maintaining a stable PUE below 1.2. The results underscore the potential value of matching cooling equipment with digital management capabilities, although the reported performance should be understood in the context of the specific facility and its operating conditions.