Key Takeaways
- Samsung Electronics (005970) is reportedly repurposing Line 2 at its NRD-K research fab in Giheung into a dedicated 2nm foundry line to boost capacity for next-generation HBM5 and custom cHBM.
- The facility will operate under a "send-fab" model, receiving wafers from other mass-production lines for specialized 2nm fabrication steps, with operations expected to begin by late 2028.
- This strategic pivot aims to address surging demand for AI infrastructure, specifically targeting the production of 2nm base dies for Nvidia (NVDA) and other major AI accelerator clients.
- Samsung's foundry business targets 100% capacity utilization by the second half of 2026, driven by a 200% increase in 2nm project orders compared to the previous year.
Strategic Pivot to 2nm Production
Samsung Electronics (005970) is shifting its manufacturing strategy by converting an R&D facility into a production-ready foundry line. The company is reportedly repurposing Line 2 at its NRD-K advanced semiconductor research campus in Giheung, South Korea. Originally intended for research and development, this line will now focus on expanding 2nm process capacity to meet the critical needs of the AI market.
The repurposed line will function as a "send-fab," a specialized operational model where the facility receives partially processed wafers from larger mass-production sites to perform specific high-end fabrication steps. This move is designed to accelerate the rollout of 2nm base dies, which are essential components for High Bandwidth Memory (HBM). Industry analysts suggest that repurposing an R&D line typically indicates a process node has reached a level of maturity and demand that justifies shifting away from pilot testing toward volume production.
Targeting HBM5 and Custom AI Solutions
The primary objective of this capacity expansion is the production of HBM5 and custom cHBM (Custom HBM). Samsung plans to apply its 2nm Gate-All-Around (GAA) technology to the base dies of these next-generation memory chips. By utilizing a 2nm logic node for the base die, Samsung aims to achieve a 50% speed improvement and significantly enhanced power efficiency compared to current HBM4E standards.
This integration of advanced foundry logic with high-performance memory is a key differentiator for Samsung against rivals like SK Hynix and TSMC (TSM). By manufacturing both the memory layers and the logic base die in-house, Samsung can offer a more tightly integrated solution for AI accelerators. The company has already secured a major strategic collaboration with Broadcom (AVGO), estimated at over $200 billion through 2030, which focuses on 2nm and below technologies for AI infrastructure.
Market Outlook and Capacity Goals
The decision to expand 2nm capacity comes as Samsung's foundry business experiences a significant recovery. The company expects its 2nm project wins to more than double in 2026. Current utilization rates for advanced nodes (8nm and below) have already reached high levels, and the company targets 100% utilization across its foundry business by the second half of 2026.
While the NRD-K Line 2 is relatively small in scale, its conversion provides a critical bridge for capacity until larger facilities, such as the Taylor, Texas fab, reach mass production. Samsung's roadmap now focuses heavily on refining the SF2 (2nm-class) family over the next three years, delaying the 1.4nm node to 2029 to ensure maximum yield and commercial competitiveness for its current leading-edge offerings.
Ed Liston is a senior contributing editor at TheStockMarketWatch.com. An active market watcher and investor, Ed guides an independent team of experienced analysts and writes for multiple stock trader publications.