SK hynix briefly tops Samsung as Korea’s most valuable company—and it’s reportedly eyeing a U.S. listing as soon as August

SK hynix briefly overtook Samsung to become South Korea’s most valuable company after a 5.6% surge pushed its market cap above $1.35 trillion, before a tech sell-off sent the shares down nearly 12.5% the next day. The company reported record 2025 revenue of 97 trillion won, up 47% year over year, and net profit of 42.9 trillion won, double last year, driven by AI-related high-bandwidth memory demand. SK hynix now controls about 58% of the HBM market and has filed confidentially for a U.S. listing, underscoring continued investor enthusiasm despite volatility.

Analysis

The market is effectively re-pricing the AI supply chain away from model makers and toward the bottlenecks that actually constrain deployment. SK hynix’s dominance in HBM implies that the next leg of AI capex still accrues disproportionately to memory and packaging incumbents, while GPU vendors like NVDA remain dependent on a small set of suppliers for system-level throughput; that raises the strategic value of the memory layer and supports sustained pricing power there. The first-order read is bullish for the whole AI stack, but the second-order effect is that supply scarcity can cap unit growth even if end-demand stays strong, making the most exposed beneficiaries the few vendors with qualified supply and the strongest pricing discipline.

The move also signals that equity performance in Korean semis has become reflexive and flow-driven, not just fundamentals-driven. Leveraged single-stock products can amplify drawdowns mechanically, which means the next 5-10% in either direction may be driven more by positioning than earnings revisions; that creates attractive opportunities around post-selloff dislocations rather than chasing the tape. For NVDA, the near-term implication is mixed: HBM concentration reduces supply optionality, but it also reinforces ecosystem dependence on Nvidia’s demand pull, so any deceleration in AI capex would likely hit memory names first and Nvidia second.

The contrarian point is that the narrative may be over-extended in the short run: when a supplier’s market cap briefly overtakes the platform customer ecosystem, expectations often outrun achievable margins once capacity expansions normalize. Over the next 6-12 months, the biggest risk is not demand collapse but a supply response from competitors and foundry partners that compresses memory pricing faster than investors expect. Separately, any U.S. listing or ETF-driven retail inflow could extend momentum, but it also raises the probability of a crowded-entry trap if global tech sentiment rolls over again.

Key Decisions for Investors

• Buy NVDA on 3-6 month weakness rather than strength; use the HBM scarcity narrative as a reason to own the customer with the best pricing power, but expect 8-12% drawdowns if Korean tech sentiment de-risks again.
• Pair trade: long NVDA / short a basket of HBM beneficiaries via Korean-listed proxies or memory exposure; thesis is that the platform owner should be less vulnerable to supply-chain margin compression over 6-12 months.
• Avoid chasing leveraged single-stock semiconductor ETFs at current levels; if used tactically, treat them as 1-3 day momentum instruments only, with hard stops given the potential for 10%+ gap risk on broad tech selloffs.
• If U.S. listing timing becomes concrete, consider fading the debut via short-dated puts or a covered-call structure on any accessible proxy; IPO/secondary enthusiasm plus existing crowded positioning creates favorable asymmetry for a post-listing mean reversion trade.
• On any further broad selloff in global tech, rotate into the most liquid memory names only after the first forced-deleveraging wave clears; the cleaner entry is 2-5 sessions after the initial shock, when flow, not fundamentals, has done the selling.