Direct:
Korea's bid to close the chip packaging gap with Taiwan and China is two stories pretending to be one. The DigiTimes headline frames it as a single race. The parallel reporting says the substrate side and the OSAT side are running on entirely different tracks, at entirely different speeds.
Substrates first, because this is where Korea is actually competitive. Samsung Electro-Mechanics (009150.KS) told its Q1 earnings call that FC-BGA demand for AI servers and data centers is running more than 50% above production capacity. Not “tight” – oversold. CEO Chang Duck-hyun said the same thing at the March AGM and again at CES 2026, where he flagged that AI server and data center demand (networking chips, power semiconductors, GPUs) could end up at 60-70% of total FC-BGA volume. SEMCO has begun formal price hike negotiations with big-tech customers. Mirae Asset revised its FC-BGA price estimates up roughly 10%. Capex is running over 1 trillion won for the second year, concentrated on the Busan and Sejong plants plus the Vietnam facility.
The pure-play OSAT story is the opposite. TrendForce's 2024 top-ten OSAT ranking has ASE Technology (3711.TW) at $18.54B (around 45% of top-ten revenue), Amkor (AMKR) at $6.32B, and JCET (600584.SS) at $5B. Hana Micron (067310.KS) came in eighth at $920M – roughly 2.2% of top-ten OSAT revenue. TrendForce was explicit: Korean OSATs cannot close the gap with Taiwanese players short-term because TSMC and ASE have a 30-year collaboration on advanced packaging that nobody else can replicate on demand. Hana Micron is investing roughly $930M through 2026, but most of that is going to legacy memory packaging in Vietnam, not 2.5D/CoWoS work.
The real Korean packaging push is happening inside the IDMs. SK Hynix (000660.KS) is putting 19 trillion won ($13B) into P&T7 in Cheongju, breaking ground in April with the wafer test line slated for October 2027 and the wafer-level packaging line targeted for February 2028, specifically for HBM. That is bigger than any independent Korean OSAT will spend this decade. Hanmi Semiconductor's (042700.KQ) 2.5D TC Bonders are being pitched at SEMICON SEA to Amkor and ASE as the named target customers, which is a useful tell about where the equipment maker sees near-term advanced-packaging volume.
For SSD readers, this is an upstream allocation signal more than a direct controller shortage call. The FC-BGA squeeze is concentrated on AI server silicon – GPUs, accelerators, server CPUs, networking – and consumer SSD controllers don't use that substrate class. But the broader AI-server-eats-everything dynamic is the same one driving the NAND and DRAM contract markets, and it's worth tracking as one piece of a bigger picture.
Sources used:
- DigiTimes, “Korea races to narrow chip packaging gap with Taiwan and China” (May 14, 2026, headline/deck only – article paywalled)
- Korea Herald, “Samsung Electro-Mechanics pours W1tr into AI substrates” (April 2026)
- Seoul Economic Daily, “AI-Driven Chip Substrate Shortage Prompts SEMCO Price Hike” (April 30, 2026)
- Korea Times, “Samsung Electro-Mechanics eyes AI boom” (CES 2026, January)
- TrendForce, 2024 top-ten OSAT rankings press release (May 13, 2025)
- TrendForce, “Korean Advanced Packaging Industry's Rise Delayed by TSMC and ASE Expansion” (July 2024)
- KED Global / Korea Times / DCD, SK Hynix P&T7 announcement coverage (January 2026)
- Seoul Economic Daily, Hanmi Semiconductor 2.5D TC Bonder at SEMICON SEA 2026 (May 2026)
Reddit: https://ift.tt/02yNSfp