MHI Targets 10-MW Centrifugal Chiller for North American Data Centers

Mitsubishi Heavy Industries (MHI) announced plans to work toward introducing a 10-MW class centrifugal chiller designed for next-generation AI data centers in the North American market, targeting industrial-grade AI computing infrastructure. The company said it aims to obtain U.S. safety and regulatory certifications, including UL, potentially as early as 2026, with commercial availability expected afterward.

As global demand for large-scale AI computing continues to accelerate, data center operators are increasingly prioritizing proven reliability, energy efficiency, and water sustainability when evaluating cooling infrastructure. MHI said it plans to leverage its industrial experience to position its centrifugal chiller solution as a potential standardized cooling platform for GW-class AI data centers, where a strong operational track record is a key factor in technology selection.

MHI’s centrifugal chillers are supported by extensive experience in industrial and mission-critical applications, including a domestic market share of approximately 60% in Japan. That track record provides a foundation for data center operators seeking dependable cooling infrastructure capable of supporting continuous, high-load operation. The proposed solution incorporates several design features aimed at improving performance and operational efficiency.

The system uses a high-efficiency centrifugal chiller design with an in-house developed compressor, optimized through a single-compressor configuration to achieve strong performance at both part-load and full-load conditions. MHI also plans to integrate plant-level control through its proprietary control system and Modular Chiller Plant architecture, enabling the effective use of free-cooling operating modes and improving Power Usage Effectiveness.

To address growing concerns around water scarcity in large-scale data centers, the system will include low-water-consumption cooling configurations using dry coolers, helping operators improve Water Usage Effectiveness. The design approach emphasizes operational stability, repeatability, and long-term efficiency, rather than reliance on unproven technologies. These features are expected to support energy efficiency and environmental sustainability while maintaining the reliability required for high-density AI workloads operating at unprecedented scale.

MHI said the newly announced 10-MW class chiller is designed to align with emerging reference architectures for large-scale AI computing facilities, which increasingly require higher cooling capacity, greater redundancy, and standardized deployment models. By addressing the cooling requirements of GW-class AI data centers, the company aims to support the evolution of data centers from traditional IT facilities into integrated AI computing environments, where cooling infrastructure plays a direct role in enabling computing performance, uptime, and total cost optimization.

Turboden’s Work

Last week, Turboden America, a subsidiary of Turboden S.p.A. and Mitsubishi Heavy Industries, was selected by Tallgrass to supply three waste heat-to-power (WHP) organic Rankine cycle (ORC) plants at gas compressor stations in Columbus, OH, Chandlersville, OH, and St. Paul, IN. The WHP units will each supply approximately 10 MW of additional clean electricity for rural utilities—recycling turbine exhaust heat into baseload, grid-stabilizing power.

The projects will also be supported by Turboden’s cloud-based AI monitoring system, Turboanalytics, to provide anomaly detection, predictive maintenance, and forecasting. These contracts qualify as waste energy recovery properties and are eligible for investment tax credit incentives of up to 50% under the Inflation Reduction Act, improving operational margins and long-term value.