Samsung's 'Kairos' Giheung fab will be a foundry plant, memory-ready

Samsung Electronics has decided that its new Giheung semiconductor facility, built under the project name 'Kairos', will be a foundry plant — with its design leaving room for later conversion to memory production. Industry sources confirmed the decision on November 4, 2026. The move signals how the AI semiconductor market is forcing manufacturers to hedge on capacity before a single wafer is produced.

What is the Kairos project?

Kairos is not an individual factory. It is the name of Samsung's project to build a new semiconductor facility at its Giheung site in Yongin, Gyeonggi Province, where the company's semiconductor business began. The site laid the foundation for the growth of Samsung's memory business. The plan now is to develop Giheung into an advanced foundry production hub.

Samsung had been reviewing medium- to long-term expansion plans at Giheung, aiming to strengthen both R&D and production. The company initially planned two buildings for a new R&D center on the site of the former General Technology Research Institute, known as SR5. It later changed course and decided on a production facility. The fab's purpose remained under discussion for some time. Options included a DRAM production line or foundry operations. Foundry won.

Why build in conversion flexibility?

The notable feature is not the foundry decision itself. It is that Samsung is designing the plant from the outset to allow conversion into a memory production facility later. Foundry and memory production differ in their processes and equipment configurations. A real conversion would demand additional investment in equipment and process adjustments. Designing for that possibility now suggests Samsung wants the option to reassign the fab as market conditions shift.

Samsung did not state a conversion timeline. It did not disclose the fab's capacity, target process node, or investment amount either. Those gaps matter. Conversion-friendly design is a promise of optionality, not of low cost.

Construction is already underway. Samsung E&A said it is carrying out structural work on the Kairos complex building, a Green Building and a substation at the Giheung site.

What is driving the decision?

The global AI semiconductor market sits behind Samsung's deliberations. Big tech investment in AI data centers is driving sharp demand growth for high-bandwidth memory (HBM, stacked DRAM for AI processors) and server DRAM. Competition to secure production capacity is also continuing in the advanced foundry market, which produces AI accelerators and other products.

Both fields matter because AI semiconductors pair high-performance memory with advanced logic chips. Samsung, an integrated device manufacturer (IDM, a company that both designs and manufactures chips) with both memory and foundry arms, is expanding capacity in line with market demand and customer investment plans. Kairos forms part of that medium- to long-term strategy — the same strategy that treats Giheung not as legacy ground but as a next-generation R&D hub with added production.

How does Kairos fit the wider Giheung plan?

Samsung plans to invest approximately 20 trillion won in NRD-K, a next-generation semiconductor R&D complex, by 2030. NRD-K covers research into next-generation technologies and product development across all semiconductor areas: memory, System LSI and foundry. With Kairos advancing alongside it, Giheung is expected to combine advanced R&D and production on one site.

A semiconductor industry official said: "As the AI semiconductor market grows rapidly, decisions about investment timing and equipment configuration are becoming as important as securing new production facilities." The official added: "Given the nature of semiconductor fabs, which require large-scale investment, a strategy that takes medium- to long-term demand into account will determine future competitiveness."

That framing cuts both ways. Flexible design buys Samsung optionality in a volatile AI market. Whether the hedge pays off depends on demand patterns the company cannot yet see.