HPE Ships First NVIDIA Grace Blackwell GB200 NVL72 System with Advanced Liquid Cooling for AI

HPE Unveils Cutting-Edge AI Infrastructure Solution

Hewlett Packard Enterprise (HPE) has announced the shipment of its first NVIDIA Blackwell family-based solution, the NVIDIA GB200 NVL72 1

1

2

2

. This rack-scale system represents a significant leap forward in AI infrastructure, designed to meet the growing demands of service providers and large enterprises for scalable, high-performance AI clusters.

Advanced Liquid Cooling Technology

A standout feature of the GB200 NVL72 is its advanced direct liquid cooling solution. This technology is crucial for optimizing efficiency and performance in the face of escalating power requirements and data center density dynamics 2

2

. HPE's expertise in liquid cooling, developed over five decades, positions them uniquely to support complex liquid-cooled environments.

Technical Specifications and Performance

The NVIDIA GB200 NVL72 boasts impressive technical specifications:

• 72 NVIDIA Blackwell GPUs and 36 NVIDIA Grace CPUs
• High-speed NVIDIA NVLink interconnect
• Up to 13.5 TB total HBM3e memory with 576 TB/sec bandwidth 2

  2

This shared-memory, low-latency architecture is designed to handle extremely large AI models with over a trillion parameters in one memory space 1

1

. The system integrates NVIDIA CPUs, GPUs, compute and switch trays, networking, and software to deliver extreme performance for heavily parallelizable workloads such as generative AI model training and inferencing 1

1

2

2

.

Industry Impact and Applications

The GB200 NVL72 is poised to address critical challenges in the AI industry. Trish Damkroger, Senior Vice President at HPE, emphasized the system's ability to offer "scalability, extreme performance, and fast time-to-deployment" 1

1

. This is particularly relevant for AI service providers and large enterprise model builders who face increasing pressure to scale their operations efficiently.

HPE's Leadership in High-Performance Computing

HPE's track record in high-performance computing is noteworthy. The company has built seven of the top 10 world's fastest supercomputers and delivered eight of the top 15 supercomputers on the Green500 list, which ranks the world's most energy-efficient supercomputers 2

2

. This experience translates into lower cost per token training and best-in-class performance for customers.

Comprehensive Support Services

To complement the hardware, HPE offers a range of support services:

• On-site engineering resources
• Performance and benchmarking engagements
• Sustainability services for energy and emissions reporting 2

  2

These services are designed to ensure rapid installation, faster time-to-value, and optimal system performance throughout the lifecycle of the infrastructure.

Future Implications

The shipment of the GB200 NVL72 marks a significant milestone in the evolution of AI infrastructure. As Bob Pette from NVIDIA noted, this collaboration between HPE and NVIDIA will "help service providers and large enterprises efficiently build, deploy and scale large AI clusters" 2

2

. This development is likely to accelerate advancements in various AI applications, from scientific research to enterprise-scale generative AI implementations.