1. Samsung Onyang Becomes a Second Base for AI Memory
Samsung Electronics broke ground on November 12, 2026, on a new high-bandwidth memory (HBM) packaging facility at its Samsung Onyang campus in Asan, South Korea. The project includes five HBM packaging lines under a 46 trillion won (roughly $34 billion) back-end expansion plan, and operations are scheduled to begin in June 2028.
The groundbreaking turns Onyang into a second major production base for advanced memory built for artificial intelligence workloads. Until now, much of Samsung's high-end packaging activity has been concentrated at its nearby Cheonan campus, and that site has been running short of room. Rather than continue squeezing more capacity out of a constrained location, the company is spreading the work across two campuses in the same region.
The decision reflects a broader reality in the semiconductor industry. Demand for memory designed to feed AI accelerators has grown faster than manufacturers can add capacity, and the bottleneck is no longer only in wafer fabrication. The final stages of production, where memory dies are stacked, connected and tested, have become just as important to how many finished chips reach customers.
2. What Is Samsung HBM and Why Does Packaging Matter?
Samsung HBM is the company's line of high-bandwidth memory, a type of DRAM in which several memory dies are stacked vertically and linked so data can move to a processor far faster than with conventional memory modules. It sits alongside the graphics processors and custom accelerators that train and run large AI models, where memory bandwidth often limits overall performance.
Packaging is the step that makes that stacking possible. After memory dies are manufactured on silicon wafers, they must be thinned, stacked, bonded together and connected through thousands of tiny vertical pathways before being tested and shipped. This back-end work is considerably more complex for HBM than for ordinary memory products, and it requires dedicated cleanroom space, specialized equipment and tight process control.
Because each HBM stack combines multiple dies, a problem at the packaging stage can affect the yield of an entire unit. That makes packaging capacity a direct constraint on how much HBM a manufacturer can actually deliver, regardless of how many memory wafers it produces upstream. Expanding back-end lines is therefore one of the most direct ways a memory maker can raise its HBM output.
3. Inside the Five-Line Onyang Build-Out
The Onyang plan centers on five HBM packaging lines. Samsung has earmarked 46 trillion won, about $34 billion at the conversion used in reporting on the project, to expand its back-end manufacturing capacity, making this one of the largest commitments the company has made to the packaging side of its memory business.
The first concrete milestone is the construction of Line 5, which is expected to bring the site's cleanroom area to roughly 31,000 square meters. Cleanrooms are the controlled environments where airborne particles, temperature and humidity are kept within strict limits, and their size is a practical measure of how much advanced manufacturing a site can host.
The campus as a whole is set to grow considerably. Total floor space at Onyang is planned to rise from 270,000 square meters to 420,000 square meters, an increase of about 55%. That additional room is aimed squarely at the space shortage that has built up at Cheonan, giving Samsung somewhere to place new HBM equipment instead of rearranging existing lines.
The schedule has also been pulled forward. Operations at the new facility are now targeted for June 2028, about one year earlier than Samsung had previously planned. In an industry where new production capacity often takes several years to come online, accelerating a timeline by a full year is a meaningful signal of how urgently the company wants additional HBM output.
4. Why Samsung's Cheonan Campus Reached Its Limits
Cheonan has been under capacity pressure for some time. As demand for HBM climbed, Samsung concentrated more of its advanced packaging work there, and the site gradually ran out of room for further expansion.
The company has responded with stopgap measures. It has shifted equipment between facilities to make space for higher-value production, and it has outsourced part of its conventional memory module work to outside partners. Those steps freed up room for HBM in the short term, but they did not solve the underlying problem of a campus that could not physically accommodate the growth Samsung wanted.
Outsourcing lower-margin products in order to dedicate in-house capacity to advanced memory is a familiar pattern among chipmakers when demand for premium components spikes. It allows a company to prioritize the products that matter most to its largest customers while keeping supply of standard parts flowing through third parties. The trade-off is reduced direct control over part of the product line, which is one reason a permanent capacity solution like Onyang carries strategic weight.
5. A Campus With More Than Three Decades of History
Onyang is far from a new site for Samsung. The campus has been operating since 1991, and for most of that time its main role has been the assembly and testing of conventional semiconductors.
That history matters for the new project. A site that already handles back-end processes has existing infrastructure, an experienced workforce and established supply relationships, all of which can shorten the path to running a more advanced operation. Converting part of a long-established assembly and testing campus into an HBM packaging hub is a different undertaking from building on an empty plot, even if the equipment and processes involved are far more sophisticated than those used for older products.
The transformation also illustrates how the industry's center of gravity has shifted. Assembly and testing were once viewed as the less glamorous end of chipmaking, but the rise of stacked memory and advanced packaging for AI hardware has turned these back-end steps into a competitive battleground.
6. How Onyang Fits Into the Wider Chungcheong Plan
The Onyang build-out belongs to a larger Samsung semiconductor program in South Korea's Chungcheong region, which also includes upgrades at the Cheonan campus. Within that program, the 46 trillion won Onyang commitment is the single largest component.
Pairing the two sites gives Samsung a regional cluster for back-end memory production. Cheonan continues to handle advanced packaging, while Onyang takes on the additional HBM lines that Cheonan cannot fit. Having both campuses in the same region can simplify logistics, staffing and coordination between facilities that are working on related stages of production.
7. South Korea's Industrial Policy Behind the Investment
The project also lines up with a South Korean government initiative built around large-scale industrial megaprojects, underscoring how central chipmaking has become to the country's economic planning.
Semiconductors are among South Korea's most important exports, and memory chips in particular have long been a pillar of the national economy. Samsung's role as one of the world's largest memory producers means that its investment decisions carry weight well beyond the company itself, affecting suppliers, regional employment and the country's position in global technology supply chains.
Aligning a corporate expansion with a government program can bring practical advantages, from coordination on infrastructure to support with permitting and regional planning. For policymakers, a flagship project of this scale serves as a visible demonstration that the country intends to remain at the front of the AI hardware race rather than cede ground to competitors elsewhere.
8. What the Expansion Means for the AI Memory Market
Supply of HBM for AI servers has been persistently tight. Reports indicate that customer demand has been outpacing the supply that memory makers have already committed under contract, leaving buyers competing for limited output.
Samsung's new capacity is designed to help narrow that gap. A June 2028 start date means the Onyang lines would begin contributing while the market is still expected to be supply constrained, which helps explain why the company chose to move the schedule forward by a year rather than keep its original timeline.
For companies building AI infrastructure, the availability of HBM shapes how quickly new accelerators can be produced and deployed. Additional packaging capacity from one of the largest memory manufacturers could ease some of that pressure over time, although the effect will depend on how fast the new lines ramp up once they begin operating.
The expansion is also relevant to the wider technology economy, including the crypto and digital asset sectors that increasingly overlap with AI computing. Data center operators, hardware makers and infrastructure providers all depend on the same constrained pool of advanced chips, so capacity decisions at the memory level ripple through costs and timelines across many industries.
9. What's Next for Samsung Onyang
Several milestones will show whether the plan stays on track. The first is progress on Line 5 construction, which will establish the initial block of new cleanroom space at the campus.
The second is whether the June 2028 operating target holds. Having already pulled the date forward by a year, Samsung will face close scrutiny over whether construction and equipment installation keep pace with that accelerated schedule.
A third question is how much of the conventional memory module work that Samsung moved out of Cheonan will remain with outside partners once Onyang is running. If the new space allows the company to rebalance its operations, some of that work could shift again, though no decision on that has been described.
Finally, the remainder of the Chungcheong program, particularly the planned Cheonan upgrades, will matter alongside Onyang. Taken together, the two campuses will largely determine how much advanced AI memory Samsung is able to ship in the second half of this decade.