
If SK Hynix and Samsung both hit their capex targets, South Korea will have committed $1.23 trillion to memory fab expansion by 2030—enough to saturate global demand and crash margins unless AI server demand sustains the 40%+ CAGR required to absorb the new capacity. This forces Micron and Intel into a choice: match the capex (which their balance sheets won't support) or cede the HBM and advanced DRAM segments to Korean suppliers. The second-order effect: US allies in the Indo-Pacific (Japan, Taiwan, Australia) will depend on Korean memory supply for military systems—a geopolitical leverage point Beijing will exploit if Seoul's strategic autonomy erodes.
SK Hynix's capital commitment directly competes with Samsung's parallel $520B state-backed investment announced three days prior, forcing Seoul's industrial policy to bet on both firms simultaneously rather than consolidate capacity.
The timing matters: both announcements land within the 72-hour window after Rocket Lab's $8B Iridium acquisition, signaling that integrated vertical supply chains (launch-to-orbit, memory-to-platform) are now a competitive requirement, not an option. SK Hynix's NAND and DRAM segregation by geography (Cheongju vs. Yongin) suggests the firm is hedging against single-site disruption risk—a constraint that didn't exist before the Taiwan contingency planning became explicit in allied procurement cycles.
What is the actual capex split between NAND and DRAM across the $712.5B total? The article cites 'only detailed investments' in those two segments—does the remainder fund packaging, testing, or new wafer starts elsewhere?
Strategic intelligence, synthesized daily — with a public track record. Every call graded against what actually happened.