SK hynix approves $38B investment for two new Korean memory fabs
SK hynix has approved a $38.3 billion investment to build two new semiconductor manufacturing facilities, including a fab dedicated to HBM and DRAM production. The infrastructure expansion is designed to support the increasing demand for high-speed memory required by data centers and artificial intelligence applications.
Key Takeaways
- Construction of the $25 billion Yongin Y2 fab will begin in July 2027, with the first cleanroom scheduled for June 2029.
- The Yongin Y2 facility spans 279 acres and will focus exclusively on DRAM and High Bandwidth Memory (HBM) production.
- A second fab in Cheongju dedicated to NAND flash is expected to open its first cleanroom in December 2028.
- SK hynix and Sandisk recently introduced High Bandwidth Flash (HBF) technology, which offers up to 16 times the capacity of standard HBM.
- The investment is part of a broader $430 billion long-term plan supported by the South Korean government to expand the domestic chip sector.
Why It Matters
This capital injection directly addresses the persistent supply-demand imbalance for high-performance memory that currently restricts the scaling of streaming recommendation engines and generative AI models. By securing the world's largest HBM production footprint, SK hynix reinforces its role as the primary supplier for high-end graphics cards and AI accelerators. For the streaming ecosystem, this expansion signals a long-term shift where memory performance, rather than just raw compute, dictates the efficiency of real-time video encoding and large-scale data processing. The industry should monitor SK hynix's HBM4 yield rates as the Yongin site ramps up, which will determine the availability of next-generation hardware for hyper-personalized content delivery.
Additional Context
The expansion comes as the global High Bandwidth Memory (HBM) market experiences unprecedented volatility. Per Counterpoint Research in June 2026, SK hynix maintains a dominant 58% share of the HBM market by revenue, though it faces intensified pressure as Samsung and Micron ramp up their own competitive architectures. Samsung, in particular, reported in early August 2026 that it is developing zHBM technology, a 3D-stacked concept designed to sit directly atop AI accelerators to achieve eight times the performance of the upcoming HBM5 standard.
Market tension is exacerbated by supply constraints that have left most top-tier memory producers sold out through 2026. According to reporting from Dow Jones in August 2026, research firm Omdia projects that annual demand for DRAM and NAND will grow by 19% through 2030, driven by the structural transformation of memory from a passive component into core AI infrastructure. This demand has significantly impacted the bill of materials for data center operators, with server DRAM prices reportedly rising between 60% and 70% as of early 2026 per industry tracking.
Furthermore, the South Korean government's involvement highlights a strategic shift toward nationalizing chip supply chains. Per WCCFTech in July 2026, the public-private initiative spearheaded by President Lee Jae Myung aims to double South Korea's memory output within five years. This includes accelerating timelines for the Yongin semiconductor cluster, which was originally slated for completion in 2045 but has been pulled forward to the early 2030s to maintain a decisive lead over rivals like TSMC and Intel in the advanced memory packaging sector.
Read full article at siliconangle.com
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source