Revolutionizing AI Data Centers: SoftBank, ZutaCore, and Foxconn's Rack-Integrated Cooling Solution

Transforming AI Data Centers with Innovative Cooling Technology

In a significant leap forward for data center efficiency, SoftBank Corp., ZutaCore®, and Hon Hai Technology Group (Foxconn®) have collaborated to develop an advanced cooling solution tailored for AI servers leveraging NVIDIA accelerated computing. This new innovation marks the debut of ZutaCore's two-phase Direct Liquid Cooling (DLC) technology in a server environment, confirmed through rigorous performance evaluations.

A Game-Changer in Cooling Solutions

On February 28, 2025, the trio announced that they have successfully implemented ZutaCore's groundbreaking two-phase DLC technology within AI servers equipped with NVIDIA's H200 GPUs. This achievement not only represents the first of its kind globally but also illustrates the urgency of improving cooling efficiency as AI-driven workloads continue to surge.

The design focuses on integrating all essential components of AI servers—especially cooling systems—into a unified rack solution. This integration facilitates an operational demonstration that verified the solution’s capabilities at SoftBank's data center, where it passed NVIDIA’s temperature standard (NVQual). Notably, the cooling efficiency was measured with a partial Power Usage Effectiveness (pPUE) of 1.03 per rack, emphasizing environmental sustainability and cost-efficiency.

The Need for Sustainable Solutions

As AI technology proliferates, the demand on data centers escalates, leading to increased power consumption and a pressing need for reduced carbon emissions. In response, this collaboration not only addresses the energy challenge but also showcases how innovative cooling technologies can significantly lower total power usage. Since May 2024, SoftBank and ZutaCore have been steadfast in developing solutions specifically aimed at enhancing the energy efficiency of AI data centers.

Innovative Features of the Rack-Integrated Solution

The novel rack-integrated solution incorporates several key features:
1. Plug and Play Capability: By streamlining the integration of server components into a singular rack solution, this technology allows for immediate usability. Users can simply connect the equipment, significantly easing installation in AI data centers.
2. Efficient Cooling Mechanism: Utilizing a specialized, insulating heat transfer fluid, the system cools critical components like CPUs and GPUs effectively. This method enhances heat transfer and reduces the required flow rate of cooling fluids, subsequently lowering pump energy usage.
3. Robust AI Compatibility and Reliability: The newly developed AI server by Foxconn, designed using NVIDIA’s H200 GPUs and optimized with two-phase DLC, outperformed expectations during NVIDIA's certification tests, affirming its stability and compatibility.
4. Versatility for High-Density Placement: The rack is engineered for high-density server placement while ensuring cooling efficiency, aligning with industry standards and facilitating safe operation.

Insights from Leaders

Hironobu Tamba, Vice President of SoftBank's Data Platform Strategy Division, stated: "As we aim to establish one of Japan's most extensive AI infrastructures, enhancing the energy efficiency of data centers is vital. Our efforts demonstrate stable operations and improved energy efficiency for high-density GPU servers."

Erez Freibach, CEO of ZutaCore, emphasized the significance of this collaboration in setting a new benchmark for energy-efficient innovations in AI data centers: "Our partnership with SoftBank and Foxconn reveals the transformative potential of waterless DLC technology in addressing the need for sustainable energy solutions."

Foxconn's representative shared enthusiasm for the collaboration, highlighting its commitment to becoming a major influence in the AI server market.

Looking Ahead

As they further develop and commercialize this groundbreaking solution, SoftBank, ZutaCore, and Foxconn plan to collaborate with additional AI server and rack design firms. Their joint mission emphasizes the crucial balance between technological advancement and environmental responsibility, paving the way for a smarter, more resource-efficient future in AI infrastructure.

Through this partnership, they aim to meet rising global demands while also contributing to a sustainable energy landscape, ultimately driving progress in AI technology without compromising the planet's health.