Industry Analysis
Chey Tae-won's build-as-fast-as-possible pledge, conspicuously absent of a start date, is textbook strategic ambiguity masking operational uncertainty. The Gwangju facility almost certainly targets HBM3E and HBM4 TSV stacking lines, pulling forward ASML EUV and Applied Materials TSV etch orders by two to three quarters while easing the memory bandwidth ceiling on next-gen AI accelerators. The missing timeline is the real signal. It points not to environmental review but to Seoul's industrial ministry calibrating equipment import quotas between Samsung and SK Hynix, a deliberate echo of the 2018 DRAM price collapse that erased roughly $40 billion in market cap. Competitively, Samsung's P4 and P5 HBM lines enter volume production in Q2 2025, and Micron's Idaho expansion is on track. If SK cannot deliver effective HBM output before Q1 2026, its share of the NVIDIA supply chain, currently near 50 percent, faces dilution below 35 percent. Twelve to twenty-four months out: ground-breaking likely mid-2026, meaningful HBM shipments by Q3 2027. The interim capacity vacuum will sustain 30 to 50 percent spot premiums on HBM, keeping memory as the binding constraint on AI infrastructure capex through 2026.
This page displays AI-generated summaries and metadata for research purposes. Original content belongs to the respective publishers.