NVIDIA Kicks Off the Next Generation of AI With Rubin — Six New Chips, One Incredible AI Supercomputer

Core Insights - NVIDIA launched the Rubin platform, featuring six new chips aimed at creating a powerful AI supercomputer that sets a new standard for AI system deployment and security at lower costs, facilitating mainstream AI adoption [2][4] - The platform utilizes extreme codesign across its components, which include the NVIDIA Vera CPU and NVIDIA Rubin GPU, to significantly reduce training time and inference costs [3][5] Innovations and Features - Rubin introduces five key innovations: advanced NVLink interconnect technology, Transformer Engine, Confidential Computing, RAS Engine, and the NVIDIA Vera CPU, enabling up to 10x lower inference costs and 4x fewer GPUs for training compared to the previous Blackwell platform [5][10] - The platform supports massive-scale mixture-of-experts (MoE) model inference, enhancing AI capabilities while reducing operational costs [5][10] Ecosystem and Partnerships - Major AI labs and cloud service providers, including AWS, Google, Microsoft, and OpenAI, are expected to adopt the Rubin platform, indicating broad ecosystem support [6][26] - Collaborations with companies like CoreWeave, Dell, and HPE aim to integrate Rubin into their AI infrastructures, enhancing performance and scalability [9][25] Performance and Efficiency - The Rubin platform is engineered to deliver exceptional performance, with the NVIDIA Vera Rubin NVL72 rack providing 260TB/s bandwidth, surpassing the entire internet's capacity [10][20] - Advanced Ethernet networking through Spectrum-6 technology enhances data center efficiency, achieving 5x better power efficiency and reliability compared to traditional methods [19][18] Market Readiness - The Rubin platform is in full production, with products expected to be available from partners in the second half of 2026, marking a significant step in AI infrastructure evolution [22][24] - Companies like Microsoft and AWS are set to deploy Rubin-based instances, further solidifying its role in next-generation AI capabilities [23][22]