Will NVIDIA Return to China's Market? Report Says New AI Chip Could Ship by Year-End

NVIDIA (NVDA-US) is planning to start limited shipments of an artificial intelligence (AI) chip to Chinese customers as early as the end of this year, targeting the rapidly growing AI inference market—though it remains uncertain whether Beijing will approve the product.

According to The Information on Thursday (the 20th), citing two NVIDIA employees, multiple Chinese clients have already placed orders for the chip. The product is a customized version of NVIDIA’s Language Processing Unit (LPU), leveraging technology licensed from U.S. AI chip startup Groq. It is designed to work alongside NVIDIA’s Graphics Processing Units (GPUs) to accelerate response times for AI chatbots.

The report states that the chip complies with current U.S. export control regulations, allowing NVIDIA to supply it to the Chinese market. However, whether Chinese regulatory authorities will permit its domestic sale remains unclear.

At the time of publication, NVIDIA had not responded to requests for comment, and the information could not be independently verified.

NVIDIA currently dominates the global market for AI model training chips but faces fiercer competition in the AI inference segment. Inference chips are primarily responsible for running trained AI models, including processing user requests for chatbots and generative AI services.

Major Chinese tech and AI firms, including Baidu (09888-HK)(BIDU-US), have already developed their own inference chips. Back in March, reports emerged that NVIDIA was preparing a China-specific AI chip compliant with U.S. export restrictions, aiming to maintain competitiveness in China’s rapidly expanding AI inference market.