NVIDIA launches Vera Rubin platform with seven new chips for scalable AI infrastructure

NVIDIA launched the Vera Rubin platform, introducing seven new chips in full production designed to support large-scale Artificial Intelligence (AI) operations. This platform is intended to advance capabilities across multiple phases of AI processing, including pretraining, post-training, test-time scaling, and real-time inference.

The Vera Rubin system addresses operational demands by integrating computing, networking, and storage technologies to create scalable infrastructure. It targets deployment in extensive AI facilities, providing resources to accommodate complex AI workloads and agentic AI applications with an emphasis on efficiency and synchronization.

The platform consolidates various components, including the Vera Central Processing Unit (CPU), Rubin Graphics Processing Unit (GPU), NVLink 6 switch, ConnectX-9 SuperNIC, BlueField-4 Data Processing Unit (DPU), Spectrum-6 Ethernet switch, and the Groq 3 LPU accelerator. Together, these elements form a coherent supercomputer system aimed at optimizing throughput and cost efficiency throughout AI model lifecycle stages.

The Vera Rubin offering comprises several key configurations: the NVL72 rack with 72 Rubin GPUs and 36 Vera CPUs supporting large-scale model training; the Vera CPU rack providing dense, liquid-cooled processing designed for reinforcement learning and agentic AI test environments; the Groq 3 LPX rack accelerating inference tasks; the BlueField-4 STX storage rack enhancing data retrieval efficiency; and the Spectrum-6 SPX Ethernet rack optimized for low-latency, high-throughput connectivity across AI installations.

Jensen Huang, founder and CEO of NVIDIA, said, “Vera Rubin is a generational leap — seven breakthrough chips, five racks, one giant supercomputer — built to power every phase of AI.” Dario Amodei, CEO of Anthropic, said, “NVIDIA's Vera Rubin platform gives us the compute, networking and system design to keep delivering while advancing the safety and reliability our customers depend on.” OpenAI CEO Secure Access Module (SAM) Altman said, “With NVIDIA Vera Rubin, we’ll run more powerful models and agents at massive scale and deliver faster, more reliable systems to hundreds of millions of people.” Timothée Lacroix of Mistral AI commented on the BlueField-4 system's contribution to scaling agentic AI efforts.

Partners and providers, including major cloud services and system manufacturers, plan to offer Vera Rubin-based products starting in the latter half of 2026. The platform forms part of a broader reference design and ecosystem intended to improve energy efficiency, system resiliency, and infrastructure scaling under continuous AI workload conditions.