Samsung Electronics has officially entered the next phase of advanced semiconductor manufacturing. The company has completed the basic design of its second-generation Samsung 2nm SF2P process node and is now actively promoting it to attract external fabless clients.
This marks a major milestone for Samsung Foundry, as the SF2P process is no longer limited to internal use. It has now been confirmed that Samsung’s upcoming AI computing chiplet platform—developed in partnership with ADTechnology, Arm, and Rebellions—will adopt the SF2P process. This is the first time this next-generation node will be used outside of Samsung’s internal System LSI division.
SF2P stands for Samsung Foundry’s “Second Generation Samsung 2nm SF2P Process.” It’s the evolution of Samsung’s original 2nm process (SF2), which is set for mass production in late 2025. SF2P builds on the existing node with enhanced performance, power efficiency, and reduced chip size.
Key Improvements over SF2:
Performance: +12%
Power Efficiency: -25%
Chip Area: -8%
These improvements make SF2P highly suitable for high-performance and low-power chip applications such as AI processors, mobile APs, and server chiplets.
Until now, Samsung Foundry had been laying the groundwork for SF2P by conducting internal tests and collaborating closely with its Design Solution Partners (DSPs). Now, with the completion of the basic architecture and design kit (PDK), the company has entered full-scale external promotion mode.
The Process Design Kit (PDK), a vital software toolkit that allows chip designers to build layouts and simulate chips using Samsung’s manufacturing rules, is now available up to version 0.9. In the semiconductor industry, version 1.0 is considered stable for full design tape-outs. That version is expected to be released by July 2025.
Also Read – Snapdragon 8 Elite Gen 2 Leaked: 4nm, 5GHz Clock, AI Boost & GPU Upgrade – Ready for 2025 Flagships?
SF2P’s momentum is also fueled by its adoption in Samsung’s AI CPU Chiplet Platform — a project jointly led by:
Samsung Foundry
ADTechnology (Korea-based DSP)
Arm (IP provider)
Rebellions (AI chip fabless startup)
This platform aims to integrate a custom CPU chiplet (designed by ADTechnology) into Rebellions’ AI processor “REBEL.” The CPU chiplet will use Arm’s Neoverse Compute Subsystems V3 as its architectural foundation and will be fabricated using Samsung’s SF2P 2nm process.
Chiplets are small functional chips that are combined into a larger package. Instead of building a complex SoC (System-on-Chip) all at once, different modules (e.g., CPU, GPU, AI cores) are manufactured separately and later connected using advanced packaging technologies. This approach:
- Reduces design complexity
- Improves scalability
- Lowers production cost for cutting-edge chips.
Initially, this AI chiplet project planned to use Samsung’s standard SF2 process. However, due to technical scheduling delays and increased maturity of SF2P, the companies decided to adopt the enhanced second-gen SF2P node instead.
Samsung’s active push for SF2P signals its broader ambition to close the gap with TSMC, particularly in the 2nm and sub-2nm space. By landing external clients early and showcasing real-world applications like Rebellions’ AI chip, Samsung is aiming to prove the maturity and competitiveness of its GAA (Gate-All-Around) transistor technology.
If successful, SF2P could:
Re-establish Samsung’s credibility as a top-tier foundry
Attract AI, HPC, and mobile chip clients away from competitors
Kick off a new era of ultra-efficient chiplet-based designs in commercial devices
Hasnain Hasan is the founder and lead writer behind this platform. With over 5 years of hands-on experience in the tech industry, he specializes in breaking down the latest trends, tools, and technologies into easy-to-understand content. A part from writing, Hasnain is actively creating and sharing tech-related content on YouTube, Instagram, and X (formerly Twitter), where he connects with a growing audience of tech enthusiasts and learner
