SKHX Extends Its Bounce as Korea's Pension Funds Buy the Chip Dip

Who's Actually Buying

The tell for this leg isn't the tape, it's who's on the bid. Korea's pension funds have now posted two straight weeks of net buying into the selloff, and SK Hynix was the single most-bought name — roughly ₩440 billion (~$298M) accumulated in July. That matters because the National Pension Service had spent most of the year *trimming* domestic equity weight; deliberately adding Hynix after it shed nearly 38% from its June 25 record reads as a call that the drawdown overshot the fundamentals.

The fast money agrees. On July 21, Mirae Asset's top-1% return traders bought SK Hynix and sold Samsung, treating Samsung's move as profit-taking and Hynix as the higher-conviction hold. When both the largest institutional pool and the best-performing retail cohort lean the same way on a name, the bounce stops looking like a dead-cat and starts looking positioned.

The Number Underneath It

The macro backdrop is doing real work here. Korea's semiconductor exports ran $22.1 billion in the first 20 days of July, up 180.6% year over year and the highest ever for the period, with chips now 40.3% of everything the country ships. That's not a sentiment number — it's HBM and DDR5 physically leaving the country to feed US and Chinese AI-server buildouts, and Hynix is the dominant supplier of the high-margin HBM slice.

That export strength is what let the KOSPI close 3.56% higher at a record 6,747.95, with SK Hynix's Seoul line up about 4%. SKHX is a perp that converts that KRW share price to USD, so a firmer won on top of a rising underlying compounds into the dollar-denominated move you're seeing on the HIP-3 book.

The July 29 Test

None of this resolves until the print. SK Hynix reports Q2 results on July 29, the first earnings since its $26.5 billion Nasdaq ADR debut on July 10 — the largest US listing ever by a foreign company. Consensus is looking for something near ₩82 trillion in revenue on operating margins in the mid-70s.

The bull case wants confirmation that 2026 HBM capacity is sold out and 2027 demand still exceeds supply. The bear case is specific: Hynix's HBM-heavy mix is locked into multi-year contracts, which caps its ability to capture the non-HBM ASP spike lifting the rest of the industry, and HBM4 mass production has already slipped. Institutions are buying the setup ahead of the answer. If the guidance confirms the supercycle, this re-rate has legs; if it flags margin pressure, the same pool that's accumulating can turn seller fast.