Midea Building Technologies presented what it calls the industry's first power-cooling hyperconverged architecture at Asia Data Centre World 2026 in Singapore, a step toward making artificial intelligence infrastructure more energy-efficient across the region.
At Marina Bay Sands, from September 29 to 30, the company showed an integrated system aimed at a growing challenge: cooling the immense computational power AI systems require without using unsustainable amounts of energy. In an announcement published October 2, Midea outlined solutions designed for the Asia-Pacific region's tropical climates and resource constraints.
The architecture brings together power systems, energy storage, magnetic bearing cooling sources and liquid cooling computer room distribution units in one infrastructure stack. It connects power supply, energy storage, cooling and liquid-cooling distribution from the earliest stages of data center construction. Midea also released a technical white paper explaining how this system-level approach offers a quantifiable, replicable reference for integrating electrical and thermal systems. That kind of technical standardization can help speed broader adoption across the region.
The Energy Efficiency Problem
Data centers consume roughly 2% to 3% of global electricity, and that figure's climbing fast as AI expands. Cooling accounts for a substantial share of that use. Traditional approaches waste enormous amounts of power. Midea's system-level approach improves the match between power supply and cooling as rack power density rises, reducing reliance on mechanical cooling and supporting improved power usage effectiveness (PUE)—a standard measure of how much energy a data center uses relative to the power it delivers to computing equipment.
The company highlighted three products. The Magnetic CDU combines a magnetic bearing cooling source with a distribution unit, so operators don't need a separate chiller and passive CDU. It cuts equipment footprint by up to 70%, shortens deployment cycles and supports a PUE below 1.2 under suitable operating conditions. The Air-Cooled Magnetic Centrifugal Chiller targets hot, water-scarce environments, where traditional water-based cooling can create resource conflicts. It delivers a coefficient of performance (COP) of up to 5.4 and reduces dependence on cooling water—an advantage in regions where water scarcity already limits development.
The Industrial-Grade CDU supplies 2.6 megawatts of cooling capacity at a 3K temperature difference. It also provides high-availability support for liquid-cooling distribution to large AI clusters and high-density racks.
Who Bears the Infrastructure Burden
Midea highlighted its strategic collaboration with Keppel on green data center infrastructure, spanning design, delivery and operation. The Gui'an Midea Cloud Data Center in China pairs Midea hardware with Keppel's digital platform, providing up to 7,654 hours of annual free cooling and a stable PUE below 1.2. Midea described it as a replicable model for green data center development in APAC's growing AI infrastructure markets.
The geographic concentration of AI infrastructure, and the energy demands it creates, raises questions about who carries the environmental and resource costs. Asia-Pacific nations increasingly host these facilities as global AI development accelerates. Without efficient infrastructure standards, regions with fewer resources must shoulder the energy and water consumption these systems demand.
Midea's support infrastructure includes what the company calls the world's first fully AI-powered chiller Lighthouse Factory, 193 local service sites across 29 countries and spare-parts centers in 34 countries. It says it's ready to support data centers across Asia Pacific. Visitors crowded the exhibition hall at Data Center World, where discussions covered AI infrastructure upgrades, data center energy-efficiency optimization and liquid-cooling technology.
Why This Matters:
AI infrastructure development doesn't happen in a vacuum. As data centers spread across Asia-Pacific to support global AI ambitions, cooling efficiency affects energy grids, water resources and carbon emissions in host countries. Midea's technology aims to separate AI growth from unlimited energy consumption, addressing a structural problem with a practical response.
Its emphasis on open technical standards, replicable models and regional service infrastructure suggests that efficient AI infrastructure requires coordination among nations and companies, not just isolated technological innovation. For regional policymakers watching these developments, the key question is whether efficiency improvements will reach everyone or remain advantages for companies and countries that can afford premium solutions. That distinction will determine whether AI's infrastructure demands can be managed democratically or concentrate environmental burdens on nations with less bargaining power.