Samsung’s advanced semiconductor nodes at 5nm and beyond drive the most sophisticated processors powering modern computing. As an integrated device manufacturer, the company focuses on both chip design and process nodes.

There are only a few companies in the world with leading-edge process technologies capable of building highly complex processors for datacenters, supercomputers, and smartphone SoCs. This dual focus on design and manufacturing distinguishes Samsung from pureplay foundries that are focused entirely on production.

The 5nm and 4nm FinFET Era

Hitting mass production in 2021, the 5nm node pushed high-performance computing (HPC) systems to deliver greater performance while consuming less power. This capability reduces the overall energy footprint as computing capabilities continue to advance. That same year saw the launch and mass production of the initial 4nm process, known as SF4E, tailored specifically to power mobile applications.

This 4nm silicon marked the final evolution of FinFET technology prior to the industry's leap to gate-all-around (GAA) architecture. Positioned at the boundary between leading-edge and mature technologies, the 4nm process leverages extensive mass production experience to ensure high yield and process stability.

Performance gains in the 4nm process are driven by the co-optimization of both transistors and interconnects. To improve performance, the 4nm process introduces additional interconnect layers with relatively larger pitch, achieving approximately a 26 percent reduction in resistance-capacitance (RC) delay. In addition, relaxed design rules in analog regions increase layout flexibility, improving both area efficiency and timing margins.

By delivering high performance, low power consumption, and design flexibility simultaneously, the 4nm process provides a strong foundation for a wide range of applications. For example, HBM4 demands extremely high bandwidth for large-scale data transfer while operating within constrained spaces. The 4nm process minimizes power loss through low-voltage operation and low-resistance interconnects to meet these thermal management needs.

The node also supports large-scale AI chip architectures, such as Language Processing Units (LPUs), which require high-density interconnects and power efficiency. Furthermore, automotive applications demand high computational performance under strict power constraints. As autonomous driving evolves from Level 3 to Level 4, the 4nm process provides a scalable foundation offering high performance per watt.

The RF market is also transitioning toward designs with increasing digital integration. The 4nm process reduces digital area and power consumption, supporting complex signal processing in next-generation communication environments like Wi-Fi 8 and 6G.

Transitioning to 3nm GAA Architecture

Pushing for massive generation-to-generation gains in performance and efficiency, Samsung Foundry completely overhauled its approach at 3nm by switching to a gate-all-around (GAAFET) transistor structure. By contrast, competitors opted to retain existing FinFET transistor structures for their 3nm nodes to ensure the technology rolled out on schedule.

Samsung's leap forward relies heavily on advanced nanosheet channels to deliver significant improvements in power, performance, and area. The company's unique GAA technology, known as Multi-Bridge-Channel FET (MBCFET), is essential for this continuing process migration. Mass production for these 3nm chips officially kicked off in the first half of 2022.

Compared to older 5nm hardware, this inaugural 3nm GAA node allows up to a 35 percent decrease in area, 30 percent higher performance, or a 50 percent cut in power consumption. In addition to these improvements, the process maturity has increased, with 3nm’s logic yield approaching a similar level to the 4nm process.

Expanding to 2nm

Samsung's SF2 stands out as a second-generation, leading-edge technology node based on MBCFET architecture. Rolling off the assembly lines in 2025, this 2nm silicon specifically targets next-generation mobile, HPC, AI, and automotive applications.

Building on the proven foundation of the first-generation MBCFET node, SF2 delivers enhanced stability for advanced computing workloads. The transition to this architecture ensures higher performance and reliability for demanding next-generation workloads.