Nvidia Denies Plans to Launch Custom AI LPU for Chinese Market
Nvidia has officially denied a report by The Information claiming the company plans to ship a custom Language Processing Unit (LPU) to China by the end of the year. The company stated it has no LPU sales in China and no such product exists on its roadmap. This denial highlights the ongoing challenges Nvidia faces in navigating U.S. export controls on advanced AI hardware while trying to maintain a presence in the Chinese market. It also underscores the competitive landscape in China, where domestic players like Huawei are gaining ground as U.S. restrictions tighten. The rumored chip was described as a derivative of the Groq 3 LPU, manufactured on Samsung's 4nm process with 512MB of on-chip SRAM and no HBM. It was reportedly designed to bypass export restrictions on Nvidia's next-generation Vera Rubin platform by working with processors already available in China.
## BACKGROUND
A Language Processing Unit (LPU) is a specialized processor designed to accelerate natural language processing tasks and large language model (LLM) inference. Nvidia's next-generation Vera Rubin platform is a high-performance AI supercomputing architecture that is subject to strict U.S. export controls, preventing its sale to Chinese customers.