NVIDIA's Vera Rubin platform is set to revolutionize the world of AI factories, with its full-scale production ramp-up. This cutting-edge technology, designed to power agentic AI, is poised to transform the way we build and operate AI systems. With a focus on performance, efficiency, and security, Vera Rubin promises to deliver intelligence at scale, marking a significant leap forward in the capabilities of AI factories.
A New Kind of Workload
One of the key aspects of Vera Rubin is its ability to handle agentic AI workloads. Jensen Huang, founder and CEO of NVIDIA, explains that agentic AI involves complex processes such as reasoning, retrieval, tool use, and response generation, all triggered by a single prompt. Vera Rubin is specifically built to manage these demanding tasks, ensuring that AI factories can handle the performance and efficiency required for the next industrial revolution.
The Power of Scale
Vera Rubin's architecture is designed to deliver 10 times the agent throughput at scale compared to the previous generation NVIDIA Grace Blackwell platform. This significant improvement in performance is made possible by the integration of various components, including the NVIDIA Vera Rubin NVL72 systems, NVIDIA Vera CPU, NVIDIA Groq 3 LPX, NVIDIA Vera BlueField-4 STX storage, and NVIDIA Spectrum-6 SPX Ethernet racks. This unified platform forms a massive AI supercomputer, capable of handling the most demanding agentic workloads.
A Global Ecosystem
The adoption of Vera Rubin is widespread, with hundreds of NVIDIA supply chain ecosystem partners across 350+ factories and 30 countries. Top system builders, infrastructure software, and storage partners are in full-scale production, including Dell Technologies, HPE, Lenovo, and Supermicro. This global collaboration is crucial to the success of Vera Rubin, ensuring its widespread availability and integration into various AI factory deployments.
Building the Fabric for Million-GPU AI Factories
To support the scale-out and scale-across AI factory deployments, Vera Rubin introduces NVIDIA Spectrum-X Ethernet Photonics, the world's first co-packaged-optics (CPO)-based switches with 200Gb/s SerDes. This technology delivers 5 times better power efficiency, 5 times longer AI uptime, and 1.3 times faster time to deployment compared to traditional transceivers. By simplifying design and freeing up more power for compute, NVIDIA's co-packaged optics networking provides the foundational fabric for million-GPU AI factories.
Secure AI for AI Factories
AI factories often process sensitive data, regulated content, and mission-critical models, making security a top priority. Vera Rubin was designed with full-stack NVIDIA Confidential Computing, ensuring a trusted execution environment at rack scale. This includes encrypting data across high-speed interconnects, providing hardware-level attestation to prevent tampering. Additionally, the NVIDIA DOCA software platform enforces advanced security across the entire AI factory, protecting data, agents, context memory, and AI inference.
Accelerating AI Factory Buildout
The NVIDIA DSX platform provides a comprehensive solution for building and operating energy-efficient AI factories. It unifies reference designs, simulation, infrastructure software, facilities, and ecosystem technologies, streamlining the deployment process. By aligning every layer of the stack, from silicon and systems to lifecycle management and multi-tenant operations, DSX accelerates the buildout of AI factories, setting a new standard for operational reliability and resiliency.
Availability and Future Prospects
Production shipments of Vera Rubin are set to begin this fall, marking a significant milestone in the evolution of AI factories. As Vera Rubin powers agentic AI factories worldwide, it will fuel the development of tomorrow's intelligence, driving innovation and transforming industries. With its focus on performance, efficiency, and security, Vera Rubin is poised to become a cornerstone of the AI revolution, shaping the future of artificial intelligence.
