Supermicro Enhances AI Infrastructure with NVIDIA Vera Rubin NVL72 Racks

Supermicro Unveils NVIDIA Vera Rubin NVL72 Racks

Supermicro, a renowned provider of complete IT solutions, has announced the launch of its new NVIDIA Vera Rubin NVL72 racks. These racks are specifically designed to support NVIDIA's cutting-edge AI technologies through seamless integration with Supermicro’s Data Center Building Block Solutions (DCBBS) and their advanced Direct Liquid Cooling (DLC-2) system. This innovative setup is expected to significantly enhance deployment speed and online performance for AI applications.

Accelerating Deployments

Supermicro's DCBBS ensures that all critical IT infrastructure components are efficiently aligned to accelerate the time-to-online for NVIDIA Vera Rubin NVL72 implementations. By consolidating cooling, power management, and integration into a single package, Supermicro enables businesses to order scalable units that are ready for deployment immediately.

According to Charles Liang, President and CEO of Supermicro, the company has dedicated years to developing the necessary components, production capacity, and deployment teams for this momentous release. Supermicro's proprietary DLC-2 technology enables it to offer a comprehensive solution for the NVIDIA Vera Rubin platform, designed to meet the intense data processing demands of AI workloads.

Maximal Cooling Efficiency

The NVIDIA Vera Rubin NVL72 racks feature a fully optimized liquid cooling stack, which is crucial for managing the high thermal output associated with AI operations. Each rack is engineered to perform as a single machine, housing 72 NVIDIA Rubin GPUs and 36 NVIDIA Vera CPUs, interconnected via advanced NVLink-switch technology, thereby providing extraordinary data throughput of 216 TB/s.

With the capability of storing up to 54 TB of high-bandwidth memory (HBM4), the cooling solutions, including in-row cooling distribution units rated at 1.8 MW, ensure efficient heat management. Optional rear door heat exchangers further assist in maintaining optimal thermal conditions within data centers.

Simplifying Integration

Supermicro's DCBBS blueprint system simplifies the process of scaling AI infrastructure. By specifying a balanced bill of materials tied to various power profiles, the DCBBS blueprint for one Vera Rubin NVL72 Scalable Unit details a configuration that includes 1,152 GPUs and 331 TB of memory spread across 16 compute racks. This meticulous design guarantees synchronized power supplies, cooling capacities, and networking infrastructure for maximum performance.

The entire implementation process, from site assessment, design, and integration to ongoing support, is managed by a dedicated Supermicro team, ensuring that projects are executed with precision and expertise.

A Leader in IT Solutions

Supermicro (NASDAQ: SMCI), headquartered in San Jose, California, is a global leader in application-centric IT solutions, focusing on markets such as enterprise, cloud, AI, and telecommunications. The company prides itself on bringing innovations to market ahead of competitors, having developed a full spectrum of products and services designed to support diverse IT workloads.

Through its commitment to green computing and eco-friendly practices, Supermicro aims to enhance total cost of ownership (TCO) while mitigating environmental impacts. All products are designed and manufactured in-house across its global facilities in the U.S., Taiwan, and the Netherlands, ensuring top-notch quality and scalability for clients worldwide.

Supermicro continues to evolve its offerings, providing clients with state-of-the-art servers, storage systems, and AI-driven solutions that align with modern technological demands. With the introduction of the NVIDIA Vera Rubin NVL72 racks, the company solidifies its position as a leading force in the digital transformation of infrastructure across industries, helping enterprises build the next-generation AI solutions needed for scalable success.