This “Boring” Country ETF Is Secretly A High-Octane AI Hardware Trade

The iShares MSCI South Korea ETF (NYSEARCA:EWY | EWY Price Prediction) is up 80% year-to-date and about 200% over the past year, which is not what a country fund is supposed to do, especially one that spent a decade as the poster child for the “Korea discount.” The reason it has done it anyway is hiding in plain sight in the holdings list. Samsung Electronics and SK Hynix together account for close to half the fund, and they happen to make the high-bandwidth memory that NVIDIA’s (NASDAQ:NVDA) Blackwell and Rubin GPUs cannot run without.

So when you buy EWY today, you are buying a Korean equity index in name and an AI memory levered trade in practice.

What EWY Is Actually For

On paper, EWY is a plain-vanilla single-country ETF. It tracks the MSCI Korea 25/50 Index, has been around since May 2000, and charges 0.6% a year. The traditional pitch was diversified developed-Asia exposure with a value tilt, anchored by the chaebols that dominate Korean industry.

The return engine works on two layers. The first is the underlying earnings of those conglomerates, which means semiconductors, shipbuilding, autos, and consumer electronics. The second, and the one driving the chart right now, is concentrated exposure to memory chips. Samsung and SK Hynix together make up close to half the fund, so EWY’s performance is tethered to DRAM pricing, HBM market share, and capex cycles in a way no S&P 500 sleeve replicates.

There is also a policy kicker. Korea’s Corporate Value-Up program is pushing chaebols to lift payouts, disclose more, and close the governance discount that has weighed on valuations for years. If it works, the rerating shows up first in the names that dominate EWY.

Why This Boring Fund Has Been Anything But

The AI thesis is concrete. NVIDIA reported $68.13 billion in fiscal Q4 revenue, with Data Center alone at $62.31 billion and networking up 263% year-over-year. Jensen Huang told investors “computing demand is growing exponentially, the agentic AI inflection point has arrived”, and guided fiscal Q1 to roughly $78 billion, excluding China. Every Blackwell and Rubin board needs HBM, and SK Hynix is the dominant HBM3E supplier with Samsung qualifying behind it.

The corroborating evidence is at Micron Technology (NASDAQ:MU), the closest US-listed comparable. Its Cloud Memory unit hit $5.28 billion last quarter at a 66% gross margin, and management is guiding next quarter to $18.7 billion in revenue with $8.42 in non-GAAP EPS. CEO Sanjay Mehrotra has said order books stretch into 2027. Micron’s stock is up 102% year to date. EWY is up 68%. NVIDIA, oddly, is up only 6%.

That spread is the whole story. The Korean memory duopoly has been catching up to a thesis the market already priced into NVIDIA. NVIDIA itself ratified the link by announcing a partnership with the Korean government, Samsung, SK Group, Hyundai, and NAVER Cloud to deploy more than 250,000 of its GPUs across Korean AI infrastructure.

The Catches You Should Know About

EWY is not a clean way to express the AI memory trade. Three things to weigh.

1. Concentration risk cuts both ways. With Samsung and SK Hynix close to half the fund, a memory pricing rollover or an HBM share loss to Micron would hit EWY harder than a typical country index. The other half is autos, shipbuilders, and banks that have their own cyclical exposures.
2. Governance is a thesis, not a guarantee. The Value-Up rerating depends on chaebols actually paying out cash and simplifying ownership structures. That is a multi-year political process, and the market has already paid for some of the optimism.
3. Currency and macro overlay. You are taking won exposure, Korean rate policy, and trade-war headlines on top of the AI thesis. Reddit chatter on EWY is still neutral with low activity, suggesting retail has not crowded in, but it also means liquidity and narrative support sit with institutions.

EWY makes sense as a 3-7% sleeve for investors who want HBM and AI memory exposure without going all-in on a single chipmaker, and who can stomach being wrong about Korean governance reform if the AI capex cycle cools at the same time.