4nm Remains Fully Loaded, Reports Say Samsung Foundry Proposes 5nm Alternative to Customers
Samsung Foundry's 4nm capacity is fully loaded due to surging orders for HBM4 and NVIDIA Groq LPU, and is now promoting the 5nm process as an alternative to customers.
Samsung Electronics' foundry business is facing capacity tightness, with its 4nm advanced process lines in a state of continuous full load. This phenomenon mainly stems from significant growth in orders for the memory business's HBM4 Base Die and NVIDIA Groq LPU-related orders.
HBM4 is the next-generation high-bandwidth memory, crucial for AI large model training and inference. The NVIDIA Groq LPU is an AI inference chip designed specifically for language processing. The surge in orders for these two types of products directly reflects the strong market demand for AI computing infrastructure.
To alleviate 4nm capacity pressure, Samsung is guiding some chip design companies to switch to the 5nm process. Since both belong to the same technology node family, this strategy helps balance capacity allocation and ensure supply stability for AI-related chips, but may also affect some customers' choices regarding process advancement.
This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.