These chip stocks could extend historic gains by rallying over 110%, according to Nomura

South Korea’s chip rally continues to be driven by AI memory demand, with SK Hynix up 183% and Samsung Electronics up 134% this year and the Kospi up 75% in 2025. Nomura sees further upside of about 117% for SK Hynix and over 110% for Samsung over the next 12 months, citing a structural memory super-cycle across DRAM, HBM and SSDs. Early earnings confirm the trend, with SK Hynix first-quarter 2026 operating profit up fivefold year on year and Samsung’s up more than 750%.

Analysis

The key mispricing is not just that memory is in an upcycle, but that the market is still treating this as a normal cyclical rebound rather than a capacity-constrained re-rating of the entire supply chain. If AI buildout continues to outgrow wafer starts, the value capture shifts away from end-demand beneficiaries and toward the few vendors with scarce process, packaging, and HBM know-how; that creates a second-order squeeze on less differentiated memory peers, module assemblers, and downstream server OEMs facing input-cost inflation. The implication is that earnings estimates are probably still too low for the leaders, but too high for everyone else in the memory ecosystem.

The next phase of upside likely comes from margin mix, not unit growth. As HBM and long-term contracts dominate mix, pricing power should persist even if spot prices flatten, which means the strongest stocks can keep compounding while the market wrongly extrapolates “peak cycle” behavior. The risk is that supply response is not linear: if capex is finally unlocked or qualification bottlenecks ease faster than expected, the air pocket in 6-12 months could be sharp because sentiment is already crowded and positioning is likely extended after the run.

Contrarian takeaway: the consensus is probably underestimating the persistence of demand from inference and agentic workloads, but overestimating the breadth of the trade. This is becoming a winner-take-most setup where the highest-quality names deserve premium multiples, while generic memory exposure is vulnerable to eventual normalization. For macro-aware portfolios, the bigger risk is not missing the upside in the leaders, but owning the wrong second-order beneficiaries that lack pricing power when the cycle eventually turns.

Key Decisions for Investors

• Go long SK Hynix exposure vs. a basket of weaker memory/module names over a 3-6 month horizon; the risk/reward favors the dominant HBM supplier capturing the tightest part of the value chain while lesser players absorb margin compression.
• Add to Samsung Electronics on pullbacks with a 6-12 month view; this is a quality-lagging catch-up trade if HBM4 qualification improves, but size it below SK Hynix because the execution gap remains the key risk.
• Short a basket of non-HBM memory and commodity server hardware names against long memory leaders as a pair trade; thesis is that input-cost inflation and mix shift will widen dispersion over the next 2 quarters.
• Use call spreads rather than outright longs if entering after the recent move; upside can persist, but implied volatility and crowded ownership make convexity preferable to naked beta exposure.
• Set a catalyst watch on any capex commentary or supply expansions over the next earnings season; if supply guidance moves up faster than expected, reduce exposure aggressively because the trade’s asymmetry will deteriorate quickly.