SoftBank Group Corp., ZutaCore and Foxconn Technology Group have successfully deployed two-phase direct liquid cooling technology in AI servers using NVIDIA’s H200 GPUs, marking the first global application of this cooling solution with these high-performance processors.
SoftBank designed a rack integration solution that incorporates the two-phase direct liquid cooling (DLC) technology among other server components. The system underwent operational demonstration and performance evaluation at SoftBank’s data center in February 2025, where it passed NVIDIA’s temperature qualification tests (NVQual), confirming its compatibility, stability and reliability.
Testing revealed the solution achieved a partial Power Usage Effectiveness (pPUE) measurement of 1.03 per rack, indicating exceptional cooling efficiency that could significantly reduce energy consumption in data centers.
The collaboration addresses growing concerns about rising power consumption as AI adoption accelerates globally. With data centers facing increased scrutiny over their carbon footprint, the industry is under pressure to develop more energy-efficient systems while managing the intense heat generated by advanced AI computing hardware.
SoftBank has been working with ZutaCore, a leading developer of two-phase direct liquid cooling technology, since May 2024 to develop low-power solutions optimized for AI data centers.
The three companies will continue their partnership to increase efficiency and reduce power consumption in AI data centers using NVIDIA’s accelerated computing technologies, with plans to develop optimized rack integration solutions for global market deployment.
