NVIDIA Vera Rubin Ramps Into Full Production to Power Agentic AI Factories Worldwide

NVIDIA’s Vera Rubin platform is now in full production, with Taiwan’s leading server makers and global supply chain partners manufacturing systems at scale to accelerate AI development. The platform is designed to power agentic AI factories, combining NVIDIA Vera Rubin NVL72 systems, Vera CPU, Groq 3 LPX, Vera BlueField-4 STX storage, and Spectrum-6 SPX Ethernet into a unified, high-performance system. This setup delivers 10x greater agent throughput compared to the previous Grace Blackwell platform, enabling large-scale AI workloads. The Vera Rubin platform marks the third generation of NVIDIA’s MGX rack-scale systems, leveraging an open-source design to streamline production. Over 150 partners in Taiwan alone, including Dell Technologies, HPE, Lenovo, Supermicro, and ASUS, are ramping up production across 350+ factories in 30 countries. These partnerships ensure widespread availability for AI labs, cloud providers, and hyperscalers. A key innovation in Vera Rubin is the NVIDIA Spectrum-X Ethernet Photonics, now in production, which integrates co-packaged optics for 200Gb/s SerDes connectivity. This technology improves power efficiency by 5x, extends AI uptime by 5x, and speeds up deployments by 1.3x compared to traditional transceiver-based networks. The platform is designed to support million-GPU AI factories with optimized performance and scalability. NVIDIA CEO Jensen Huang emphasized the platform’s role in advancing agentic AI, which requires complex reasoning and tool integration. Vera Rubin is positioned as the engine for the next industrial revolution in AI, delivering the performance, efficiency, and security needed for large-scale deployments. The ramp-up reflects NVIDIA’s strategy to dominate the AI infrastructure market through hardware and ecosystem partnerships.