Saturday, August 8, 2026
DarkSubscribe
AI Infrastructure · News & Analysis
HomeChips & HardwareReport
Chips & Hardware · Report

SK Hynix announces $38 billion investment in new AI semiconductor manufacturing fabs in South Korea.

Major capex addresses HBM memory shortage; reduces hyperscaler supply constraints and memory-side bottleneck for AI acceleration.
Trade pressSlicast · August 8, 2026 · US · Source: Google News
importance 92

SK Hynix is making one of its biggest bets yet on the artificial intelligence boom, announcing plans to invest 54 trillion Korean won, or about $38.1 billion, to expand its memory chip manufacturing capacity. The South Korean semiconductor giant said it will build two new fabrication plants dedicated to next-generation memory products, reinforcing its long-term strategy to remain a leading supplier of chips powering AI data centers and high-performance computing systems.

The investment includes 35.2 trillion won for the Y2 fabrication plant in Yongin and 19.1 trillion won for the M17 facility in Cheongju. Together, the projects are designed to increase production of dynamic random access memory, or DRAM, and NAND flash memory—two products that have become critical components in the AI industry.

Soaring demand for high bandwidth memory, or HBM, has created a supply crunch across the semiconductor industry. HBM chips are essential for AI accelerators produced by companies such as Nvidia, enabling faster data processing and improved performance for large language models, cloud computing, and advanced AI applications. That demand has fueled a sharp increase in memory chip prices over the past year, benefiting the industry's largest manufacturers. Investors have rewarded leading suppliers including SK Hynix, Samsung Electronics, and Micron Technology with strong stock gains, betting that tight supply conditions will continue for several years.

Despite its leadership in HBM technology, SK Hynix faces mounting competition in the broader memory market. According to Counterpoint Research, Samsung reclaimed the top position in the global DRAM market by share during the second quarter, intensifying pressure on its domestic rival to expand production capacity.

Neil Shah, vice president of research and co-founder of Counterpoint Research, said the new investment is aimed at securing SK Hynix's long-term manufacturing footprint rather than boosting immediate production. "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. He added that expanding investments across the industry are unlikely to ease pricing pressure anytime soon.

The Yongin Y2 facility represents the second of four fabrication plants planned for the company's Yongin Semiconductor Cluster, a major manufacturing hub intended to anchor SK Hynix's future memory production. Construction is scheduled to begin in July 2027, with the first cleanroom expected to open in June 2029. The site will manufacture HBM alongside other advanced DRAM products. Cleanrooms are highly controlled manufacturing environments that minimize contamination during semiconductor production, making them essential for producing increasingly complex memory chips.

The Cheongju M17 facility will focus on NAND flash memory, another segment experiencing rising demand as AI workloads require greater storage capacity. Groundbreaking is scheduled for February 2027, while the first cleanroom is expected to begin operations in December 2028. SK Hynix said the investment reflects both the rapid pace of AI adoption and the need to ensure sufficient production capacity for future customer demand.

The new spending represents the next phase of SK Hynix's broader expansion strategy unveiled last year. Under that long-term plan, the company intends to invest a total of 600 trillion won in developing the Yongin Semiconductor Cluster and another 100 trillion won to expand its production base in Cheongju.

Read the original
SK Hynix announces $38 billion investment in… · Slicast