Samsung to Invest $34 Billion in HBM Packaging: New Onyang Facility Aiming for 2028
The race for AI dominance is no longer just about who can manufacture the smallest DRAM chips, but who can most efficiently stack, connect, and test them. Recognizing this shift, Samsung Electronics has announced a massive expansion of its advanced packaging capabilities. On November 12, 2026, the South Korean giant unveiled plans for a new production facility at its Onyang campus in Asan, part of a broader strategic push to secure its position in the high-bandwidth memory (HBM) market. The total investment is expected to reach 46 trillion KRW (approximately $34 billion USD) across several development stages.
The scale of this investment is particularly notable due to its accelerated timeline. Samsung aims to have the new production line operational by June 2028, roughly a year earlier than previously anticipated. This move is a direct response to the skyrocketing demand for HBM, which is essential for modern AI accelerators. The goal is to transform the existing Onyang site from a semiconductor assembly and testing hub into a specialized center for advanced HBM manufacturing.
Expanding the Onyang Campus Infrastructure
The project, which saw its groundbreaking ceremony attended by Samsung Chairman Lee Jae-yong and South Korean President Lee Jae-myung, is part of a larger national initiative to bolster semiconductor and AI infrastructure. The financial breakdown reveals that the 46 trillion KRW is not a single lump sum for one building; rather, 6 trillion KRW is allocated for the initial construction of the new packaging facilities, while the remaining 40 trillion KRW is earmarked for the phased installation of cutting-edge manufacturing equipment. This new facility will serve as the fifth HBM production line at the site, spanning three floors with a total cleanroom area of approximately 31,000 square meters.
This expansion will significantly increase the Onyang campus's footprint, growing from its current 270,000 square meters to 420,000 square meters—a 56% increase. According to reports from the Yonhap news agency, the project is also expected to create around 700 direct jobs in the region.
The Complexity of HBM Packaging
To understand why this investment is so critical, one must look at the fundamental difference between standard DDR5 modules and HBM. While traditional memory places chips on a large PCB, HBM involves stacking multiple DRAM dies vertically. These layers are connected through Through-Silicon Vias (TSV)—vertical electrical connections that pass through the silicon to enable high-speed data exchange between the stacked layers. These stacks are then integrated into a single package alongside a processor or AI accelerator using complex technologies like silicon interposers. This architecture provides massive interface widths and high data rates while reducing energy consumption per bit compared to traditional external connections.
However, as memory density and stack counts increase, the mechanical and thermal challenges grow exponentially. Each die must be precisely aligned, reliably connected, and rigorously tested. Even a minor defect in a single connection within the stack can render the entire package useless, significantly impacting yields. This is why Samsung's focus on advanced packaging is just as vital as its DRAM manufacturing; the ability to scale the assembly and testing processes is the real bottleneck in meeting HBM demand.
Market Context and Uncertainties
The HBM market is currently one of the most technically demanding sectors in the semiconductor industry, with Samsung, SK hynix, and Micron competing to meet the rigorous requirements of AI accelerator manufacturers. For Samsung, success depends not just on chip production, but on packaging technology, yield rates, and successful qualification by customers who require tight integration with their specific processors. While reports from TrendForce suggest the new facility may be designed to support upcoming generations like HBM4E, official details regarding specific production volumes or the exact generation assigned to each line remain unconfirmed.
There is also some discrepancy in the projected timelines. While Samsung's presentation points to a June 2028 start for the new line, some reports, including those cited by TrendForce, suggest that mass production might not begin until May 2029. This difference likely accounts for the gap between the initial commissioning of the production line and the achievement of full-scale, high-volume manufacturing capacity.
While the scale of the investment is massive, Samsung has not provided a binding confirmation regarding the exact completion date. Currently, June 2028 remains the officially communicated timeline for the facility's commissioning. For the HBM (High Bandwidth Memory) market, this means the new site will not provide any additional memory volume in the short term. Even with a successful production launch in 2028, the manufacturer must first establish stable manufacturing processes, achieve sufficient yields, and complete the qualification process for the intended products.
The long-term economic impact of this move remains tied to the evolving demand for AI accelerators. An expansion of this magnitude requires that the investments justify themselves over a prolonged period through high production volumes and consistent customer orders. At this stage, Samsung has not disclosed specific capacity targets or expected utilization rates for the new facilities.
Strategic Context: Packaging as the New Bottleneck
The investment in Onyang represents a logical strategic shift. The development of HBM is no longer solely dependent on faster DRAM chips; without sufficiently powerful packaging facilities, it is becoming economically difficult to implement the increasing capacities and increasingly complex memory stacks required by modern computing. The scale of the planned investment in manufacturing equipment—approximately 40 trillion Won—highlights the significant effort Samsung anticipates for the next stages of expansion.
It is important to distinguish these investments from existing production capacity. The capital is being deployed in stages, and the actual ramp-up is still several years away. This accelerated timeline serves primarily as a signal to the market and potential customers. Whether Samsung can effectively improve its competitive position against SK hynix and Micron will ultimately depend on manufacturing yields, the technical quality of its HBM products, and successful customer qualification. While additional factory floor space is a prerequisite, it is not a competitive advantage in itself.