South Korean chipmaker plans new DRAM and NAND capacity as tight supply and soaring AI infrastructure spending push memory prices higher
SK Hynix said on Friday it will invest 54 trillion Korean won ($38.1 billion) to build two new memory chip manufacturing plants in South Korea, stepping up capacity expansion as surging demand from artificial intelligence infrastructure drives a shortage of key memory components.
The South Korean chipmaker will invest 35.2 trillion won in a new fabrication plant known as Y2 in Yongin and another 19.1 trillion won in an M17 facility in Cheongju.
Register for Tekedia Mini-MBA edition 20 (June 8 – Sept 5, 2026).
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
The investment comes as memory prices have surged amid tight supply and rapidly increasing demand from companies building AI data centers and chipmakers such as Nvidia, whose advanced AI processors require large quantities of high-bandwidth memory (HBM).
HBM has become one of the most strategically important components in the AI hardware supply chain because it enables AI accelerators to process large volumes of data at high speeds. The explosive expansion of AI infrastructure has therefore created a sharp increase in demand for advanced memory, while manufacturers have struggled to bring new capacity online quickly enough.
The resulting supply imbalance has benefited the world’s largest memory producers, with SK Hynix, Samsung Electronics and Micron seeing their shares rally as investors bet that strong AI demand will keep memory prices elevated for years.
The investment also comes as SK Hynix faces intensifying competition from Samsung.
Samsung reclaimed the top position in the global DRAM market by revenue share in the second quarter, according to Counterpoint Research. DRAM, or dynamic random-access memory, is widely used in computers, servers and AI systems and remains one of the industry’s most important memory products.
Neil Shah, vice president of research and co-founder of Counterpoint Research, said the competitive pressure was encouraging SK Hynix to increase capital expenditure.
“This has prompted SK Hynix to inject fresh capex to expand its footprint. In the near term, this won’t alter SK Hynix’s output but is built for 2029 and beyond,” Shah told CNBC.
The new plants will not materially increase supply immediately. Instead, they are designed to provide capacity toward the end of the decade, when AI infrastructure is expected to require substantially more memory.
Shah said capacity additions by Samsung, SK Hynix, Micron and China’s CXMT could significantly expand global memory supply through 2028, but he expects demand to grow even faster.
“Looking at the broader market, multi-vendor expansions from Samsung, SK Hynix, Micron, and CXMT will expand global supply significantly through 2028. Yet with demand growing even faster than planned capacity, memory prices are unlikely to soften before the end of 2028,” he said.
SK Hynix said the Yongin Y2 plant will be the second of four planned fabs in its Yongin Semiconductor Cluster and will serve as a production base for DRAM. Construction is scheduled to begin in July 2027, with the first cleanroom expected to open in June 2029. The facility will manufacture HBM and other next-generation DRAM products.
A cleanroom is a highly controlled manufacturing environment designed to limit contaminants that could damage semiconductor production.
The second facility, Cheongju M17, will focus on NAND memory, which is primarily used for data storage in devices ranging from smartphones and computers to enterprise storage systems and data centers.
SK Hynix said demand for NAND is also increasing rapidly as AI workloads generate and process increasingly large volumes of data.
Construction of M17 is scheduled to begin in February 2027, with the first cleanroom expected to open in December 2028.
“In the AI era, technological competitiveness alone is not enough and the ability to supply the required volume at the exact moment customers need it is the ultimate competitive advantage,” SK Hynix said.
“We reached this investment decision after a thorough review of market demand.”
The projects represent the latest phase of SK Hynix’s broader semiconductor expansion strategy. The company announced a long-term investment plan last year under which it intends to invest 600 trillion won in the Yongin Semiconductor Cluster, and 100 trillion won to expand its Cheongju production base.
The scale of the latest investment underscores how the AI boom is reshaping semiconductor capital spending. Memory manufacturers are increasingly having to balance the risk of overbuilding capacity against the prospect that AI data centers, advanced computing and memory-intensive workloads will continue to drive demand faster than new factories can supply it.
SK Hynix’s investment also represents a bet that the current AI-driven memory cycle will persist well beyond the immediate surge in HBM demand. By building capacity that will come online from 2028 and 2029, the company is taking a position for a market in which AI accelerators, servers and data centers are expected to require substantially greater memory bandwidth and storage capacity.
But some industry analysts believe the challenge for the company will be maintaining its technological lead while ensuring that the enormous capital commitments translate into sufficient returns if memory supply eventually catches up with demand.