Samsung Electronics has started full-scale mass production of Nvidia’s Grok3 Language Processing Unit (LPU). Nvidia announced the production deal on August 24 at the Hot Chips 2026 conference. Samsung utilizes its 4-nanometer process to manufacture the advanced AI inference chip.
Manufacturing yields for the foundry division have reportedly surpassed 80%. The Grok3 LPU functions as a primary component of Nvidia’s next-generation Vera Rubin AI platform. This hardware specializes in accelerating AI agent workloads and response generation.
Samsung also supplies HBM4 memory for Nvidia’s Rubin GPUs. The deal reinforces Samsung’s position as a comprehensive supplier for AI hardware.