Samsung Foundry extends mature 8nm node with 8LPU+ aimed at automotive and edge AI chips

Samsung launches 8LPU+ for automotive and edge AI

Samsung Foundry has launched 8LPU+, a second-generation extension of its 8nm process aimed at automotive, robotics and edge AI chips that do not need the most advanced nodes. As reported by Automotive World, Samsung says the new process delivers 12% more performance, 15% lower power consumption and an 8% smaller logic area than the previous generation. ¹

Industry reporting carried by Sina Finance's market desk, published a day before Automotive World's coverage, says Samsung Foundry will introduce the 8LPU+ process platform within the year, citing the same headline figures: a 12% performance improvement, 15% lower power and a logic area reduction of about 8% compared with 8LPU. Both accounts frame the process as targeting chips where the cost of leading-edge nodes such as 2nm is not justified. ¹⁶

The performance and power figures are Samsung's own claims for the process generation; no independent benchmark results for 8LPU+ silicon appear in the published material, and the timing of volume availability beyond the stated launch window has not been detailed. ¹⁶

Why a mature DUV node, not EUV

8LPU+ runs on existing deep ultraviolet (DUV) production lines, so Samsung and its customers avoid the cost of extreme ultraviolet (EUV) lithography, according to Automotive World's report of the Samsung Foundry announcement. The 8nm node is the last DUV process generation in Samsung's advanced lineup before EUV insertion. ¹⁷

The economics are central to the pitch. Sina Finance's reporting notes that EUV machines cost hundreds of billions of Korean won each, are supplied by a single vendor, ASML, and push wafer costs higher, whereas 8nm can be produced on fully depreciated DUV equipment with well-established yield data, predictable costs and existing automotive and industrial chip qualifications, lowering the supply barrier for carmakers and tier-one suppliers. ⁶

Automotive World characterises the move as Samsung competing for automotive business where cost and reliability matter more than transistor density, at a time when the foundry is working to keep pace with TSMC at the leading edge. That competitive framing is the publication's own analysis rather than a Samsung statement. ¹

Zonal architectures drive the target chips

The process is aimed at carmakers moving to more centralised electronic architectures, which need more processing power to control functions across the vehicle, and much of that work can be done without 2nm-class silicon, per Automotive World's account of the launch. ¹

Sina Finance identifies the core target products as AI microcontrollers (MCUs) and automotive zonal controllers (ZCUs). Under the zonal architecture now common in the industry, the vehicle is divided into zones, each with its own controller covering doors, body functions and powertrain, which increases chip demand per vehicle. ⁶

For these applications, Sina Finance's industry sourcing says the decisive metrics are not peak compute but operating temperature range, long-term supply continuity, embedded storage and unit price, since automotive and industrial robot MCUs and zonal controllers must work reliably in harsh environments. Samsung's own automotive materials stress the same priorities, citing packaging that withstands thermal extremes, electromagnetic interference and constant vibration over an expected 15-year vehicle life. ²⁶

Technical refinements: I/O, leakage and eMRAM

Sina Finance's detailed report describes several area and power refinements on the 8LPU+ platform. The process integrates 3.3V planar I/O devices on 8LPU FinFET wafers, cutting I/O area by about 10%, which matters because even as edge AI compute cores shrink, input/output circuitry and embedded non-volatile memory (eNVM) occupy substantial die area. ⁶

The process also addresses standby leakage when a vehicle is switched off or a robot idles, through ultra-low-leakage SRAM (ULL-SRAM) and high-threshold-voltage devices (cHVT). The magnetic embedded memory eMRAM on the platform reaches automotive grade 1 (AG1) specification, with a cell area of 0.017 square microns and storage density of roughly 20Mb per square millimetre, allowing more storage in the same area and smaller controller chips. ⁶

These specifications come from Sina Finance's industry-sourced account of the Samsung announcement; Samsung's own specialty technology pages confirm it markets FinFET-based eMRAM with high-temperature reliability and high density as a cost-effective non-volatile memory option, but do not republish the 8LPU+ specific cell figures in the material reviewed here. ⁵⁶

A decade-long 8nm automotive platform

8LPU+ extends a node with history. Samsung announced 8LPP (8nm Low Power Plus) in May 2017 as the most competitive scaling option before transitioning to EUV, combining process innovations from its 10nm technology, according to the company's own roadmap announcement at the Samsung Foundry Forum that year. ⁷

Per Sina Finance's account, the 8nm process entered commercial use after 8LPP was developed in 2017, with the performance- and density-optimised 8LPU following in 2018; 8LPU+ is the second-generation refinement on that base. Automotive World notes the node already has an automotive track record, since Nvidia's Drive Orin platform is built on Samsung's 8nm process. ¹⁶

Samsung's automotive foundry materials confirm 8nm logic is offered at Auto Grade 2 (AG2), with plans to expand auto grade support to AG1 for select processes including 8nm, alongside 14nm at AG1 and 5nm (SF5A) at AG2, with 4nm and 2nm automotive nodes under development. Samsung's design methodology for these platforms integrates AEC-Q100-aware sign-off, ISO 26262 functional safety certification and ASIL-B assessment readiness, with the automotive service package certified to IATF 16949. ²

The eMRAM element of 8LPU+ was foreshadowed in Samsung's own 2023 strategy. At the Samsung Foundry Forum 2023 in Munich, the company said it would expand its eMRAM portfolio from 28nm FD-SOI and 14nm to 8nm by 2026 and 5nm by 2027, and that its 8nm eMRAM showed potential for 30% higher density and 33% higher speed than the 14nm process. ³

What it means for chip supply into vehicles

Sina Finance's reporting suggests the likely customer base includes existing 8nm clients within Samsung's foundry ecosystem and fabless chip companies seeking automotive qualification, noting that domestic and overseas design houses and ASIC vendors already have experience taping out edge AI system-on-chip designs using 8nm eMRAM, lowering adoption barriers. That customer outlook is attributed to an industry source rather than a Samsung statement. ⁶

Samsung's broader automotive portfolio shows the 8nm node sits alongside other manufacturing options: the company mass-produces in-vehicle infotainment solutions on 14nm, 8nm, 5nm and 4nm processes, and has announced automotive silicon ranging from Exynos Auto processors made on 5nm to 130nm BCD power processes, with dedicated 8-inch PMIC wafers for power management. ²⁴

For buyers of vehicles, the significance is indirect: zonal electrical architectures depend on affordable, automotive-qualified controllers, and extending a proven, cost-predictable node is aimed at keeping those chips economic. The reported launch is of the process platform itself; no specific customer chip designs, production volumes or vehicle programmes for 8LPU+ have been announced in the material published so far. ¹⁶