NVIDIA Vera Rubin NVL72 Rack

Executive Summary

The NVIDIA Vera Rubin NVL72 is a rack-scale AI supercomputer designed to power agentic reasoning AI and the AI industrial revolution. It unifies 72 Rubin GPUs and 36 Vera CPUs to deliver high-performance computing capabilities. With its advanced architecture and design, the Vera Rubin NVL72 is poised to revolutionize the field of artificial intelligence.

The Vera Rubin NVL72 is a significant development in the field of AI computing, offering unparalleled performance and efficiency. Its unique design and architecture make it an attractive solution for businesses and organizations looking to deploy large-scale AI infrastructure. With its ability to deliver 3.6 EFLOPS of NVFP4 inference and 2.5 EFLOPS of training compute, the Vera Rubin NVL72 is a powerful tool for driving innovation and growth in the AI industry.

Architecture & Design

The Vera Rubin NVL72 features a unique architecture that combines 72 Rubin GPUs and 36 Vera CPUs in a single liquid-cooled rack. The Rubin GPUs are designed to handle AI workloads, with each GPU featuring 288 GB of HBM4 memory and 22 TB/s bandwidth. The Vera CPUs, on the other hand, are designed for data center workloads, with each CPU featuring 88 custom Arm-based Olympus cores and up to 1.5 TB of LPDDR5X memory.

The Vera Rubin NVL72 also features a range of other components, including the Groq 3 LPU, NVLink 6 switch, ConnectX-9 SuperNIC, BlueField-4 DPU, and Spectrum-6 switch. These components work together to provide a high-performance and efficient computing platform. The Groq 3 LPU, for example, is designed to handle decode-phase inference acceleration, while the NVLink 6 switch provides a high-bandwidth interconnect between the GPUs and CPUs.

The Vera Rubin NVL72 is designed to be highly scalable, with a full POD scaling to 40 racks, 1,152 GPUs, and 60 exaflops. This makes it an attractive solution for businesses and organizations looking to deploy large-scale AI infrastructure. The platform is also designed to be highly reliable, with advanced features such as proactive maintenance and real-time health checks.

Performance & Thermal

The Vera Rubin NVL72 is designed to deliver high-performance computing capabilities, with 3.6 EFLOPS of NVFP4 inference and 2.5 EFLOPS of training compute. The platform is also designed to be highly efficient, with a power consumption of not publicly disclosed. The thermal design of the platform is also not publicly disclosed.

Despite the lack of publicly disclosed thermal design and power consumption specifications, the Vera Rubin NVL72 is designed to be highly reliable and efficient. The platform features advanced cooling systems and power management capabilities, making it suitable for deployment in a range of environments.

Market Positioning

The Vera Rubin NVL72 is positioned as a high-end AI computing platform, designed to meet the needs of businesses and organizations looking to deploy large-scale AI infrastructure. The platform is priced at not publicly disclosed, making it a significant investment for any organization. However, the platform's high-performance capabilities and advanced features make it an attractive solution for businesses looking to drive innovation and growth in the AI industry.

The Vera Rubin NVL72 is expected to be deployed by a range of businesses and organizations, including AWS, Google Cloud, Microsoft, OCI, and CoreWeave. These deployments are expected to take place in H2 2026, with the platform being made available to partners in Q1 2026.

Verdict

The NVIDIA Vera Rubin NVL72 is a powerful and efficient AI computing platform, designed to meet the needs of businesses and organizations looking to deploy large-scale AI infrastructure. With its unique architecture and advanced features, the platform is poised to revolutionize the field of artificial intelligence. While the platform's pricing and thermal design are not publicly disclosed, its high-performance capabilities and reliability features make it an attractive solution for businesses looking to drive innovation and growth in the AI industry.

Overall, the Vera Rubin NVL72 is a significant development in the field of AI computing, offering unparalleled performance and efficiency. Its unique design and architecture make it an attractive solution for businesses and organizations looking to deploy large-scale AI infrastructure. As the platform becomes more widely available, it is likely to have a significant impact on the AI industry, driving innovation and growth in a range of fields.