en.Wedoany.com Reported - July 14 - According to reports, Nvidia and Japan's Mitsubishi Heavy Industries (MHI) are planning to collaborate in the field of artificial intelligence data center technology. The two parties are considering integrating MHI's cooling systems and energy management technologies into Nvidia's next-generation data centers. This collaboration is still in the research and negotiation phase, with no formal agreement, specific projects, construction locations, or technology implementation timelines announced yet.
The proposed partnership will focus on addressing the heat dissipation and power supply challenges posed by high-density AI servers. As the power density of GPU clusters continues to increase, data center infrastructure needs to simultaneously adjust server cooling, room power distribution, and energy management methods. Nvidia's new generation of AI server cooling solutions, announced in June, allows coolant to operate at temperatures up to 45 degrees Celsius, reducing the operational load of peripheral refrigeration equipment such as chillers by raising the coolant temperature.
Japan's Mitsubishi Heavy Industries has developed a data center technology portfolio covering on-site power generation, power supply, efficient cooling, and digital energy management. The company previously showcased on-site power generation, efficient cooling, and next-generation power distribution systems at Nvidia's GTC 2026, positioning these facilities as supporting infrastructure for high-density AI data centers.
Cooling systems may become a key area of technical integration between the two parties. MHI is advancing immersion cooling and two-phase direct-to-chip cooling technologies. The two-phase solution circulates an insulating refrigerant through a cold plate on the processor surface, directly removing heat generated by the GPU from the chip. In December 2025, MHI and Japan's EXEO Group built and commercially deployed a GPU server system using two-phase direct-to-chip cooling in a data center.
In terms of energy management, MHI announced on July 9 a cooling system optimization trial conducted at Fujitsu's Akashi Data Center in Japan. The system provides unified control over cooling equipment from multiple suppliers, reducing cooling energy consumption in the test room by 2.3% without interrupting data center operations. If expanded to the entire data center, it is expected to reduce electricity consumption by up to 7.6%. These capabilities can be used to coordinate server loads, air conditioning equipment, and cooling source systems, but it is currently unclear whether they will be directly incorporated into Nvidia's data center solutions.
If the collaboration materializes, Nvidia can set heat dissipation, power supply, and rack operation requirements from the server and GPU architecture side, while MHI may provide supporting technologies from the cooling facility, power distribution system, and energy control levels. However, at this stage, this collaboration cannot be described as the cooling system having received orders or begun deployment. Specific details on whether immersion cooling, direct-to-chip cooling, or other solutions will be adopted still await formal announcements from both parties.
