NVIDIA announced the launch of its Rubin platform, marking a major leap forward in AI infrastructure. Named in honor of pioneering astronomer Vera Rubin, the platform integrates six new chips-the NVIDIA Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU, and Spectrum-6 Ethernet Switch-to deliver one of the world's most advanced AI supercomputers. Rubin aims to accelerate mainstream AI adoption while drastically reducing training and inference costs.
The Rubin platform leverages extreme codesign across hardware and software, providing up to 10x reduction in inference token cost and enabling the training of large-scale mixture-of-experts MoE models with 4x fewer GPUs compared with the NVIDIA Blackwell platform. Advanced innovations include NVIDIA NVLink interconnect technology, Transformer Engine, Confidential Computing, and the RAS Engine, all designed to optimize performance, security, and reliability.
Performance and Efficiency BreakthroughsRubin introduces next-generation AI-native infrastructure, including the NVIDIA Inference Context Memory Storage Platform, powered by BlueField-4. This enables scalable agentic AI reasoning with efficient data sharing and power-efficient performance. The platform's Spectrum-X Ethernet Photonics switches deliver 5x improved power efficiency and uptime, while BlueField-4's Advanced Secure Trusted Resource Architecture ASTRA ensures secure and isolated AI workloads.
