Samsung Foundry is set to extend its 8nm process, which has been in existence for over eight years, for yet another generation. Observers interpret this move as a strategy to efficiently utilize existing mature deep ultraviolet (DUV) production lines while avoiding the costly extreme ultraviolet (EUV) lithography equipment.
The process is aimed at automotive, robotics, and edge AI chips that do not require the high costs associated with 2nm-class nodes. According to industry sources on the 20th, Samsung Foundry will launch its second-generation 8nm 8LPU+ process platform within the year. Compared to the previous 8LPU, it delivers a 12% performance improvement, a 15% reduction in power consumption, and an approximately 8% reduction in logic area.
Samsung's 8nm process was first commercialized in 2017 following the successful development of 8LPP, and it represents the last DUV (argon fluoride immersion lithography) node among Samsung's advanced processes before the introduction of EUV. In 2018, Samsung introduced 8LPU, which focused on performance and integration density optimization. The newly announced 8LPU+ is a second-generation version further refined on the basis of 8LPU.
Current leading-edge processes such as 5nm, 4nm, and 2nm offer greater computing power, but critical lithography layers must use EUV equipment. A single EUV machine costs hundreds of billions of Korean won, driving up wafer costs, and only ASML globally can supply it, limiting equipment availability. In contrast, the 8nm process can be produced on DUV equipment lines that have already been fully depreciated. With long-term accumulation, yield data is sufficient, costs are predictable, and automotive and industrial chip certifications have already been obtained, lowering the barrier to supplying automakers and automotive component suppliers.
The core target products for 8LPU+ are expected to be AI microcontrollers (MCUs) and automotive zonal controllers (ZCUs). The automotive industry has recently widely adopted the ZCU architecture, which divides a vehicle into multiple zones, each with its own controller. Under this architecture, each zone covering doors, body, and powertrain requires controllers, driving up chip demand.
MCs and ZCUs used in automotive and industrial robotics need to operate stably in harsh environments. Rather than peak computing power, operating temperature range, long-term continuous supply capability, embedded memory, and unit price are more critical metrics. Chip miniaturization is also a key focus. Even if the computing core of an edge AI chip is shrunk, the I/O responsible for external connectivity and the embedded non-volatile memory (eNVM) that retains data after power-off still occupy significant area.
The 8LPU+ platform integrates 3.3V planar I/O devices on an 8LPU FinFET wafer, successfully reducing I/O area by approximately 10%. The process also addresses standby leakage current issues when a vehicle is turned off or a robot is in standby mode, featuring ultra-low leakage SRAM (ULL-SRAM) and high-threshold-voltage devices (cHVT). The magnetic memory eMRAM, which retains data after power-off, meets the automotive grade 1 (AG1) specification, with a cell area of 0.017μm² and a memory density of approximately 20Mb/mm², allowing more storage in the same area and helping further reduce controller chip size.
An industry source said: "Existing 8nm customers within Samsung Foundry's ecosystem and fabless chip companies seeking automotive certification will both be potential customers for 8LPU+. Domestic and overseas design companies and ASIC vendors already have experience tape-out of edge AI system-on-chips (SoCs) based on 8nm eMRAM, so the adoption barrier is relatively low."
Comments