On November 12, 2026, Samsung Electronics held a groundbreaking ceremony for a new advanced packaging plant for high-bandwidth memory (HBM) at its Onyang campus in Asan, South Korea. According to ChosunBiz, the plant's start-up target has been moved up by about a year from the original plan, to June 2028. Samsung plans to invest a total of 46 trillion won in stages to build five HBM production lines at Onyang, and the groundbreaking moves the site's buildout into the full-scale construction phase. According to the conversion Yonhap News gave as of November 12, the investment is equivalent to about 5.53 trillion yen.
The investment is notable for expanding capacity in the stacking and interconnection steps that are essential to making HBM. HBM is not finished simply by manufacturing DRAM dies; multiple dies must be stacked and electrically connected. Even if the new plant starts up earlier, HBM shipments will not necessarily rise immediately. Actual supply capacity depends on how far production equipment has been installed and on yields once mass production begins.
About 87% of the 46 trillion won total goes to production equipment
Of the 46 trillion won planned for Onyang, 40 trillion won is earmarked for production equipment, about 87% of the total. This is a ratio calculated from the 46 trillion won total and 40 trillion won equipment cost reported by Yonhap News on November 12; the remaining 6 trillion won is for construction. In other words, the plan puts far more money into the manufacturing equipment that will go inside than into the buildings themselves. These figures are a breakdown of the staged investment plan covering Onyang's five lines, not an amount spent solely on the newly started plant.
According to Yonhap News, the new Line 5 plant, on which construction has now started, will have three floors and about 31,000 square meters of cleanroom space. With the expansion, the Onyang campus's total floor area is expected to grow about 1.6 times, from the current 270,000 square meters to 420,000 square meters. However, the new plant's cleanroom area and the total floor area of the campus including existing facilities are different measures and cannot be compared directly.
Meanwhile, Asia Business Daily's on-site report puts the investment in expanding cleanroom production facilities at Onyang at 1.3 trillion won. That figure covers a different scope from the 46 trillion won total for the five Onyang lines and the 6 trillion won construction portion within it. The 1.3 trillion won was reported as the investment in expanding production facilities, and it would not be appropriate to simply add it to the total.
The groundbreaking ceremony was held alongside the fourth public-private review meeting that the South Korean government had scheduled for November 12. Yonhap News and ChosunBiz also reported on the on-site ceremony, and the plan to expand Onyang's production capacity has moved into the construction phase. However, as the large share of equipment spending shows, once the building is completed, manufacturing equipment must still be moved in, installed, and the production processes tuned. Progress on construction and on equipment installation therefore needs to be tracked separately.
Separately, in the investment plan Samsung announced on June 29, combined investment in Cheonan and Onyang was put at 56 trillion won, and group investment across the entire Chungcheong region at 140 trillion won. These differ in geographic scope and announcement date from the 46 trillion won for Onyang alone, so they need to be treated separately. To judge how much Onyang's supply capacity will grow, what matters is not only the total investment but also what equipment will be installed on the five lines and when they will start operating.
Expanding stacking and interconnection is essential to raising HBM output
HBM is high-bandwidth memory made by stacking multiple DRAM dies vertically and electrically connecting them with through-silicon vias (TSVs). Samsung's HBM product materials also describe this stacked structure. In addition to the front-end process of making dies on semiconductor wafers, as with ordinary DRAM, it requires an advanced packaging process that stacks and connects multiple dies into a finished product. The South Korean government's announcement also positions the new Onyang plant as an advanced packaging facility.
To increase HBM supply, both the capacity to make DRAM dies and the throughput of the process that stacks and connects them must be secured. Even if the front-end process can produce large numbers of dies, shipments of finished products will not rise if packaging throughput falls short. Conversely, adding stacking and interconnection equipment will not be fully utilized if the required number of DRAM dies cannot be secured. The Onyang investment aims to expand advanced packaging, the part of the HBM manufacturing flow that governs finished-product supply capacity.
In next-generation HBM, not only does the number of stacked DRAM dies increase, but the logic base die at the base of the stack also plays an important role. Under the HBM4 specifications Samsung has published, 1c-process DRAM is combined with a 4nm-process logic base die in a 12-layer configuration with 36GB of capacity. The per-pin data transfer rate is up to 13Gbps.
For the next-generation HBM4E, Samsung plans to use 1c DRAM and a 4nm logic base die while extending to up to 16 layers and up to 64GB. It targets a per-pin data transfer rate of up to 16Gbps and memory bandwidth of up to 4TB/s for the whole stack. Capacity indicates how much data can be stored, the per-pin rate indicates how fast data moves through each individual signal pin, and stack-wide bandwidth indicates the data transfer capability of the product as a whole. These are different metrics, and more layers or greater capacity do not by themselves mean faster transfer speeds. Moreover, these are product specifications published by Samsung, not mass-production results achieved on the new Onyang lines.
According to ChosunBiz, Onyang's Line 5 is planned to produce next-generation products such as HBM4E, which corresponds to seventh-generation HBM. Adding layers can expand HBM capacity, but it also raises the difficulty of thermal management and manufacturing. Samsung also explains the thermal countermeasures and manufacturing challenges that come with more layers. Mass-producing next-generation products requires manufacturing technology that can connect more dies with high precision while securing sufficient yields.
Developing a product that achieves the maximum specifications of HBM4E and the like is therefore a separate challenge from producing it stably in volume. The 40 trillion won in planned equipment investment at Onyang is intended not just to add manufacturing equipment but to expand production capacity for these complex stacking and interconnection processes. Supply of next-generation HBM will depend heavily on equipment throughput and on yields at mass production.
Start-up target moved up to June 2028; mass production expected from May 2029
Asia Business Daily reports that the new Onyang plant will start operating in June 2028 and that HBM mass production will begin in May 2029. What ChosunBiz describes as moved up by about a year from the original plan is the plant's start-up date. It would therefore be inaccurate to read June 2028 as the start of full-scale HBM mass production. Nor does this start-up target mean completion of the entire 46 trillion won Onyang investment plan or simultaneous operation of all five lines.
Behind Samsung's push to expand are growing demand for AI memory and constraints on existing production capacity. In its third-quarter 2026 results announced on October 29, the company explained that, with capacity limited, it is shifting its product mix toward AI products and significantly expanded HBM4 sales. The approach is to raise the allocation of existing capacity to AI products while expanding next-generation HBM supply capacity at Onyang.
For 2027 as well, the company forecasts that capital spending on AI infrastructure will reach record levels and that AI-related demand will far exceed industry supply capacity. It also indicated a policy of building a stable sales base through multi-year contracts. Given this demand outlook, bringing forward the new plant's start-up can be seen as an effort to get capacity in place quickly for future HBM demand.
However, the percentage increase in plant area cannot be taken as the HBM output increase rate. Onyang's total floor area is planned to grow about 1.6 times, but actual production capacity will vary with the number and performance of the manufacturing equipment installed, the throughput of the stacking and interconnection processes, and finished-product yields. For products with many layers, such as HBM4E, securing stable yields while coping with more complex manufacturing processes will be especially important.
Going forward, the focus will be on progress in installing equipment toward the June 2028 start-up target and on building out the production system before mass production, reported for May 2029. Once monthly output and yields after start-up become known, it will be easier to assess how much actual supply capacity the 46 trillion won investment produces. Whether Samsung can stably source the DRAM dies it needs from the front-end process and secure sufficient throughput and yields in stacking and interconnection will be the key to turning the Onyang expansion into a full-scale increase in next-generation HBM supply.