The project has a name inside Samsung: Kairos. According to a report by the Korean outlet fnnews on Nov. 3, it is centered on advanced foundry capacity at the company’s Giheung campus, south of Seoul, and it includes an assessment of whether some memory lines at the site can be converted to other uses. The work sits around the Giheung NRD-K research complex, the campus where Samsung develops next-generation process technology.
The detail that gives the project its shape is a decision Samsung had already made. The company has said it will shut all of its 8-inch foundry lines by the end of 2030. Those lines make the older, larger-geometry chips that go into displays, power management and automotive systems. Closing them frees floor space, power and staff at a campus that has been running since the 1980s, and the question the Kairos project appears to be answering is what fills the space afterward.
The answer, on the evidence available, is advanced logic. That is a harder conversion than it sounds. An 8-inch line and a leading-edge 300-millimeter line share a building and almost nothing else. The plumbing for ultra-pure water, the gas handling, the vibration isolation, the cleanroom classification and the number of process steps all differ. Converting a memory fab to logic is harder still, because memory and logic use different tool sets and different yield models, and the people who run one are not interchangeable with the people who run the other.
Samsung is nonetheless one of the few companies with practice at the maneuver. It has spent a decade moving capacity between DRAM, NAND and logic at its Korean sites as markets turned, and it has the engineering depth to attempt a conversion that a pure-play foundry would struggle to justify. The cost, in capital and in lost output during the transition, is the reason the strategy is rare.
The location explains the timing. Giheung is where Samsung’s research and development sits, and NRD-K is the complex built to develop the processes that Hwaseong and Pyeongtaek later run at volume. A conversion project placed around a research campus is not a volume play in its first phase. It is a place to prove that a converted line can hit yield before the company commits a larger fab to the same design.
The foundry angle is where the commercial stakes sit. Samsung’s contract chipmaking business has spent years trying to close the gap with the market leader, and its problem has never been capacity in the abstract. It has been qualified capacity: lines that customers have tested and trusted with their designs. Every additional qualified line is a bargaining chip in negotiations with the large chip designers who split their orders between two suppliers to keep prices down.
Adding capacity in Korea also reduces a geographic risk that customers have started to price. A foundry customer choosing between suppliers now asks where the wafers will be made and what happens if that place becomes difficult to operate from. Korean capacity is easier to sell than it was five years ago, when the same customer might have insisted on a second site in another country.
The memory conversion carries a second message. Samsung is the largest maker of memory in the world, and its memory lines have been the source of both its profits and its volatility. Moving some of that floor space to logic is a statement that the company expects the mix of demand to change, and that it would rather own a smaller share of a stable business than a larger share of a cyclical one. It is the kind of decision that looks obvious in hindsight and is expensive to make in advance.
The 8-inch business is the part of the portfolio that has become a commodity. Those lines serve customers who need modest volumes and mature processes, and they compete with Chinese fabs that have added capacity steadily and priced aggressively. Closing them removes a low-margin drag from the income statement without touching the leading-edge business that Samsung’s strategy depends on.
The question the conversion project addresses is what to do with the buildings once the tools are gone. A fab shell is the most expensive part of a semiconductor site and the least portable. Demolishing one and starting over is a way to spend money twice. Converting it, if the process can be made to work, is a way to spend it once.
The report does not say how much Samsung intends to spend, how many lines would be converted, or when the first converted line would produce wafers. Those omissions matter, because a conversion announced and a conversion completed are separated by two to three years of tool installation and process tuning in the best case. Samsung has not confirmed the project publicly.
What the report does establish is direction. A company closing its 8-inch foundry lines by 2030 while studying whether memory lines can be repurposed is a company reallocating its oldest assets toward its newest ambition. Giheung was where Samsung learned to make chips at all. The Kairos project, if it proceeds as described, would make the original campus the place where the company tests whether it can still learn something new.
