[Jensen Huang Explains the Purpose of Vera Rubin: A New Heterogeneous Computing Foundation for the Agent Era] According to monitoring by Beating, Jensen Huang, during his GTC Taipei 2026 keynote, redefined the design logic of the Vera Rubin platform around the computational needs of Agents, emphasizing that 'Vera Rubin is not a single chip, not just a GPU, but a fully end-to-end designed system.' He described it as the most ambitious project in NVIDIA's history, involving 40,000 engineers across the company. Huang explained the necessity of such a system through the operational mechanism of Agents. At its core, an Agent is an aggregated and distributed heterogeneous computing model: large language models (the brain) think on GPUs, with each inference activating an entire row of Grace Hopper NVLink 72; the orchestration engine (the body) schedules the entire workflow on CPUs; tool invocation (the workshop) utilizes both CPUs and GPUs simultaneously; the security layer runs on NVIDIA BlueField DPUs, enabling end-to-end encryption of data at rest, in transit, and in use, adhering to confidential computing standards. The memory system encompasses KV caching, data compression, and retrieval, which Huang predicted would revolutionize the storage industry. The Vera Rubin full stack includes: Rubin GPUs + NVLink 6 interconnect, Vera CPUs, ConnectX-9 SuperNICs, BlueField-4 DPU security processors, and the DOCA software stack. Huang also announced that all CUDA X libraries will be equipped with 'skills' (Agent Skills), enabling Agents to learn how to call these libraries as if reading a manual. In the future, Agents will use CUDA X with efficiency far surpassing that of humans. Huang stated that NVIDIA has evolved from a GPU company into a systems company and is now undergoing another transformation: customers no longer want to buy computers; they want to build AI factories. NVIDIA's ecosystem has expanded to include power supply, cooling systems, electrical grids, and other industrial infrastructure, with the goal of delivering a complete full-stack system that allows customers to directly construct AI infrastructure. [Original Article Link]