Key facts
- Nvidia and Mitsubishi Heavy Industries are exploring a potential partnership.
- The collaboration focuses on integrating Mitsubishi's cooling and energy management systems into Nvidia's AI data centers.
- The goal is to improve energy efficiency and thermal management for AI infrastructure.
- Mitsubishi Heavy Industries has expertise in power generation, refrigeration, and industrial machinery.
- Nvidia is increasingly referring to its next-generation AI data centers as 'AI factories'.
Nvidia and Mitsubishi Heavy Industries are reportedly in discussions for a partnership that could see the Japanese industrial conglomerate supply advanced cooling and energy management systems for Nvidia's artificial intelligence data centers. This potential collaboration aims to tackle the escalating electricity consumption and heat generation challenges associated with the rapid expansion of AI technologies.
The partnership would leverage Mitsubishi Heavy Industries' extensive experience in areas such as power generation, gas turbines, aerospace, industrial machinery, and large-scale refrigeration and air-conditioning equipment. These capabilities are seen as directly applicable to the demanding operational requirements of AI data centers, which require continuous, reliable power and cooling infrastructure.
Nvidia, which increasingly describes its advanced AI data centers as 'AI factories,' is seeking solutions to manage the immense heat produced by high-performance GPUs and other components. As AI accelerators consume significant power, with modern GPUs drawing around 1 kilowatt each and server racks potentially requiring hundreds of kilowatts, conventional air cooling methods are becoming insufficient. The industry is shifting towards liquid cooling technologies, including direct-to-chip, cold plate, two-phase, and immersion cooling, alongside high-efficiency chillers and heat recovery systems.
By combining Nvidia's computing technologies with Mitsubishi Heavy Industries' thermal and energy expertise, the companies aim to enhance the efficiency and reduce the power consumption of future AI data centers. This move underscores the growing importance of cooling infrastructure as a fundamental component of AI system architecture, moving beyond a secondary consideration to a critical element for performance and competitiveness.