Samsung breaks ground on Onyang HBM packaging hub as Cheonan runs out of room

Samsung Electronics has started building a new high-bandwidth memory (HBM) packaging facility at its Onyang campus in Asan, South Korea. The company held the groundbreaking on November 12, 2026.

The reason is refreshingly physical. Its nearby Cheonan campus, where much of this work happens today, is running out of space, so Samsung is turning Onyang into a second major production base for advanced AI memory.

What Samsung is actually building

The plan calls for five HBM packaging lines. Samsung has earmarked a total of 46 trillion won, around $34 billion, for the build-out, which is aimed squarely at expanding its back-end capacity.

The footprint numbers are substantial. Construction of Line 5 is expected to push cleanroom space at the site to approximately 31,000 square meters.

Total campus floor space is set to grow from 270,000 square meters to 420,000 square meters. That works out to roughly 55% more room, which is a fairly direct answer to the Cheonan squeeze.

Timing matters too. Operations at the new facility are slated to begin in June 2028, which is described as one year earlier than previously expected.

Why Cheonan hit its limits

Cheonan has been under capacity pressure for a while. The strain has led Samsung to shift equipment around and to outsource some of its conventional memory module work.

Onyang is not a new location. The site has been running since 1991, mostly handling conventional semiconductor assembly and testing.

The Onyang project also sits inside a wider plan. Samsung has announced a total investment of 60 trillion won for its broader Chungcheong semiconductor initiative, which includes upgrades at the Cheonan campus.

So the 46 trillion won Onyang commitment is the largest slice of a bigger regional push, not a standalone bet.

The policy backdrop

The investment lines up with a South Korean government initiative called the “Three Great Leap Mega Projects.” The program underscores how central chip manufacturing has become to the country’s economic planning.

What this means for the AI memory market

The core issue is supply. HBM for AI servers has been persistently tight, and shortages are projected to continue through 2028.

Reports also indicate that customer demand is outpacing current contract supply into 2028.

Samsung’s new capacity is designed to address exactly that gap. With operations targeted for June 2028, the new lines arrive right around the period when shortages are expected to persist.

Key markers to track include progress on Line 5 construction, whether the June 2028 operational target holds, and how much of Cheonan’s outsourced conventional work stays outsourced once Onyang comes online.

Also worth monitoring is how the rest of the 60 trillion won Chungcheong program unfolds, particularly the Cheonan upgrades. Together, the two campuses will shape how much advanced AI memory Samsung can actually ship later this decade.