NVIDIA CEO Jensen Huang announced at the GTC Taipei keynote today that the Vera Rubin system, a Pod-level supercomputer designed for the era of AI agents, has entered mass production. Huang stated that NVIDIA has 150 supply chain partners in Taiwan, working together to reshape computing for the AI era, and expressed his gratitude to Taiwan. Vera Rubin is designed for AI agents to observe, reason, plan, and use tools. The system utilizes TSMC's 3nm process and CoWoS packaging, along with HBM4 memory from Micron, SK Hynix, and Samsung. Partners such as Foxconn and Quanta are involved in the assembly. Huang emphasized that the extreme co-design with Taiwanese partners has pushed the scale and efficiency of AI factories to their limits.

FACT BOX

  • Source: CNA (Central News Agency)
  • Category: product_announcement
  • Organizations: NVIDIA