South Korea's KOSPI hits record as SK Hynix joins $1 trillion club after Samsung, Micron

SK Hynix topped $1 trillion in market value for the first time, with shares jumping as much as 14.9% to lift its valuation to a record 1,680 trillion won ($1.12 trillion). The AI-driven memory chip rally also pushed Samsung above $1 trillion earlier in May and Micron above the milestone on Tuesday, as tight supply and strong demand for high-end memory chips continue to drive prices and profits higher. The KOSPI hit a record high, rising as much as 5.09%, underscoring broad market impact from the semiconductor surge.

Analysis

The market is starting to price memory like a quasi-infrastructure layer for AI rather than a cyclical component business. That re-rating is not just about better pricing; it reflects a supply regime where lead times, capex discipline, and HBM mix are now more important than unit growth, which means the earnings power of the leaders can stay elevated longer than classic semiconductor investors expect. The second-order winner is the broader AI capex stack: if memory margins remain this rich, hyperscalers and accelerator vendors face a higher bill of materials, but the cost is still small relative to the productivity gains, so demand should remain resilient in the near term.

The bigger near-term market impact is positioning, not fundamentals. With retail and leveraged ETF flows now chasing Korean semis, the trade has become self-reinforcing and vulnerable to sharp air pockets if volatility spikes or if there is any sign of pricing inflection. In that setup, the first place to look for a negative catalyst is not outright demand collapse but a subtle change in commentary around HBM allocation, customer pre-buying, or capex guidance from the memory leaders.

For MU and NVDA, the asymmetry is different. MU should continue to benefit from the same structural pricing regime, but it is more exposed to a later-cycle margin disappointment if investors extrapolate current spot pricing too far into FY26. NVDA is only marginally exposed through component costs, but any sustained memory tightness can become a narrative overhang if it compresses OEM margins or delays platform adoption at the edge. UBS’s higher targets tell us sell-side is still underestimating duration of the cycle; the contrarian risk is that consensus is now crowding into the same bullish endpoint just as the trade becomes technically extended.

The cleanest way to express the view is to own the beneficiaries with the strongest operating leverage while fading the most crowded momentum expression. The opportunity is not in calling the memory cycle over; it is in separating fundamental compounding from flow-driven overshoot. If the market continues to reward scarcity, the winners will be those with the best HBM mix and balance sheet flexibility, but if sentiment cools, the unwind can be violent because index and leveraged ETF ownership has become a meaningful marginal driver.

Key Decisions for Investors

• Maintain/ ավել weight MU on 3-6 month horizon via outright long or call spreads: still has room to rerate on structural memory pricing, but size below full conviction because the stock is increasingly exposed to a late-cycle multiple reset if forward pricing rolls over.
• Pair long MU / short NVDA for 1-2 quarters: expresses the memory upcycle directly while hedging some AI-platform crowding; risk/reward improves if AI capex remains strong but memory economics keep tightening.
• Use any 5-8% pullback in Samsung/SK Hynix-linked exposure to add only tactically, not chase strength: the trade is flow-supported now, but the marginal buyer is retail-levered and vulnerable to a 2-3 day reversal.
• Buy downside protection on crowded semiconductor beta through short-dated put spreads on broad AI/semis exposure into the next event window: the key risk is a sentiment air pocket, not a fundamental collapse, and the payoff is best in a volatility shock.
• For longer-term accounts, prefer leaders with HBM and capex discipline over broad semiconductor baskets: the next 12 months likely reward scarcity and pricing power more than volume growth.