Nvidia(US:NVDA) Ge Long Hui·2026-01-06 06:42Core Insights - NVIDIA's CEO Jensen Huang announced the new Rubin platform at CES 2026, which is set to revolutionize AI computing with significant performance improvements over the previous Blackwell platform [3][6][9]. Group 1: Rubin Platform Overview - The Rubin platform will feature six independent chips, including the core Rubin GPU, and is designed to drastically reduce inference costs by up to 10 times compared to the Blackwell platform [6][7]. - The training of mixed expert models (MoE) will require four times fewer GPUs than the previous generation, enhancing efficiency in AI model training [8][10]. - The platform is named after astronomer Vera Rubin and is expected to be delivered to initial customers in the second half of 2026 through partnerships with companies like Dell, HPE, and Lenovo [6][11]. Group 2: Open Source Models - Huang introduced four open-source models targeting different verticals: Nemotron for logical reasoning, Cosmos for understanding physical laws, Alpamayo for autonomous driving, and Clara for healthcare applications [12][13][14][16]. - The rise of open-source models is seen as a catalyst for global innovation, with several emerging models from China, such as Kimi K2 and Deepseek R1, showing competitive performance [19][20][22]. - Huang emphasized that open-source models are narrowing the performance gap with leading AI models, transforming the AI landscape and encouraging widespread participation [22].