Nvidia Groq3 LPX enters mass production as Samsung yields top 80%

Nvidia’s next-generation Groq3 LPX artificial intelligence inference accelerator has entered full-scale mass production at Samsung Electronics’ foundry business, with production yields recently rising above 80% from about 33% in April. The level is generally viewed as the threshold for stable mass production of advanced chips. Developed using technology from Groq, an AI chip company Nvidia indirectly acquired last year, the Groq3 LPX is designed to accelerate answer generation by AI agents that repeatedly perform complex computations. It is produced on Samsung Electronics’ 4-nanometer process at the S5 line in the company’s Pyeongtaek campus in Gyeonggi Province. Analysts say Samsung’s experience producing high-bandwidth memory base dies on a similar process has strengthened its manufacturing of Nvidia’s core AI chips. Nvidia CEO Jensen Huang disclosed in March at GTC 2026 that Samsung was making the chip, while Samsung later displayed Groq3 LPX 4nm wafers at its exhibition booth. Samsung’s Pyeongtaek 4nm lines have maintained high utilization amid rising AI-chip demand, and the company is reported to have increased 4nm foundry prices by up to 15%.

The information on this website is generated using AI and we cannot guarantee its accuracy. Please use it as reference information only.