- The KOSPI is up about 97% in 2026 and printing record highs — but Samsung and SK Hynix now make up more than half the index, and the two of them did almost all the work.
- On the record-high session, nine of ten KOSPI stocks actually fell, and foreign investors have net-sold for more than 14 straight sessions. Domestic retail and ETFs are holding the index up.
- The obvious Korea ETF does not diversify you out of it: Samsung and SK Hynix are roughly 48% of EWY, because the index it tracks is that concentrated.
The KOSPI record high looks like a clean win. Korea’s benchmark closed at 8,476 on May 29, up roughly 97% in 2026 — one of the best runs of any major market on earth. Look one layer down, though, and the rally is not what the headline number says it is. Two semiconductor names, Samsung Electronics and SK Hynix, now account for more than half of the entire index, and they have done nearly all of the lifting. Strip them out and “Korea is up 97%” becomes a much quieter, much narrower story.
SK Hynix — ticker 000660.KS on the Korea Exchange — changed hands at ₩2,333,000 going into publication on May 30, 2026, and is up about 245% this year. That is not a typo. A single memory-chip maker has more than tripled in five months, and because of how the KOSPI is built, its move counts for far more than the other 900-plus companies in the index combined. Understanding why that is true, and why it should make you cautious rather than excited, is the whole point of this piece.
What the KOSPI record high actually hides
The KOSPI is capitalization-weighted: a company’s influence on the index equals its market value, not one-company-one-vote. That is normal — the S&P 500 works the same way. What is not normal is the degree of concentration. As of late May, Samsung and SK Hynix together crossed 50% of the total KOSPI market cap for the first time. Inside the blue-chip KOSPI 200, the pair is about 51.5% — Samsung near 28%, SK Hynix near 23.5%. At the start of 2026 that combined weight was 38.7%. It has jumped nearly 13 percentage points in five months.
So when SK Hynix runs 245% and Samsung runs roughly 147% in a year the index gains 97%, the arithmetic is doing exactly what you would expect: two enormous, fast-rising weights drag a cap-weighted average up, even if most of the names underneath are flat or falling. The index is not lying. It is just measuring something much narrower than “the Korean stock market.”
The cleanest tell sits in the products built to remove this exact distortion. Equal-weighted Korea ETFs — which give every holding the same vote — have badly lagged the cap-weighted KOSPI this year. When the equal-weight version trails the cap-weight version by that much, it is mathematical proof that the gains live in the biggest names and nowhere else.
Nine of ten stocks fell on the record day
Here is the statistic that should stop you. On one of the record-setting sessions, nine out of ten KOSPI stocks declined while the index printed an all-time high. Decliners outnumbered gainers on a day the headline said “record.” That is only possible when a tiny handful of mega-weights rise enough to overwhelm a broad sea of red.
Narrowing breadth at new highs is one of the oldest cautionary signals in markets. It is not a sell trigger — narrow markets can stay narrow for a long time, and plenty of healthy bull runs have had ugly breadth for stretches. But it changes what the index is telling you. A broad advance says “the economy and corporate earnings are lifting everything.” A two-stock advance says “one theme is lifting two stocks, and the index happens to own a lot of them.” Those are very different risk profiles wearing the same number on the screen.
The foreigners are selling into the record
If the index is at records, someone is buying. It is not the people you might assume. Foreign investors have been net sellers of Korean stocks for more than 14 straight sessions, unloading on the order of ₩50 trillion — roughly $16 billion — even as the KOSPI made new highs. Sustained foreign selling into strength is the kind of divergence that pays to watch: it can be benign rotation, or it can be distribution into retail hands near a top. The tape does not tell you which in real time.
So who is the marginal buyer holding the index up? Domestic money. Korean retail investors and the local financial-investment sector — increasingly through ETFs — have absorbed the foreign supply. Korea just saw the debut of single-stock leveraged ETFs tracking Samsung and SK Hynix directly, which pours even more concentrated, leveraged retail demand into the same two names. The structural read: the KOSPI’s price is now set at the margin by domestic flows and product launches, not by the global funds that historically anchored it. That makes the index more reflexive — more likely to move hard on its own internal flows — and more hostage to two tickers’ daily swings.
Why it is these two names
None of this is random. Samsung and SK Hynix are the world’s two largest makers of high-bandwidth memory (HBM), the stacked, ultra-fast memory that sits beside every Nvidia AI accelerator. HBM is the genuine bottleneck of the AI build-out — you cannot ship an AI server without it — and the same demand pull that drives Nvidia runs straight through Seoul. SK Hynix lit the latest leg by unveiling “iHBM,” an integrated-cooling design for next-generation HBM5 packages. This is the Korean expression of the same trade we mapped across Asia’s handful of AI-concentration names — and it rides the broader memory pricing cycle that ultimately decides both companies’ earnings.
That is the bull case in one sentence: if AI compute keeps scaling, the HBM duopoly prints money, and a cap-weighted Korea index full of them goes along for the ride. It is a real, fundamentally-grounded story — not a meme. The catch is the flip side of any duopoly bet: when your index is the duopoly, the memory cycle is no longer one input among many. It is the only input that matters.
Bulls see KOSPI 10,000. The peak-out crowd sees the same chart.
The optimism is not shy. Samsung Securities lifted its KOSPI target to 11,000 — fueling open talk of “KOSPI 10,000” — on the back of memory super-cycle upgrades and price-target hikes for the two giants. If you believe HBM demand runs for several more years, the math is not absurd: keep compounding two names that are half the index and the index follows.
But the peak-out warnings are looking at the identical picture and drawing the opposite lesson. The rally’s health is now entirely a function of one cycle not rolling over. A wobble in memory prices, a delay in the next HBM generation, or a single disappointing capex signal from the hyperscalers would not just dent two stocks — it would take the index, because the index is those two stocks. Concentration cuts both ways: it amplified the upside on the way here, and it removes the cushion that breadth normally provides on the way down.
What it means if you want to own Korea
Most investors outside Korea reach for EWY, the iShares MSCI South Korea ETF, with around $24 billion in assets and a gain of about 100% in 2026. Here is the trap: EWY inherits the same concentration. Samsung and SK Hynix are about 48% of the fund. It even tracks a capped index — MSCI Korea 25/50, which is specifically designed to stop any single name exceeding 25% and to keep all the big positions from summing past 50%. The two chip names are pinned right against that ceiling. In other words, even an index methodology built to limit concentration can no longer dilute this particular bet. Buying “Korea” through EWY is, to a first approximation, buying levered HBM exposure with a Korean wrapper.
If you do want the names themselves, use the native Korea Exchange listings — Samsung at 005930.KS and SK Hynix at 000660.KS, priced in won — rather than the thin over-the-counter ADRs, which barely trade and track poorly. And size the position for what it actually is: a concentrated semiconductor-cycle bet, not a diversified country allocation.
The bottom line
The KOSPI record high is real on the screen and hollow underneath. Korea’s index and Korea’s market have decoupled: one is at all-time highs, the other had nine of ten stocks falling on the same day. The single variable that closes that gap — or blows it wide open — is the HBM cycle that Samsung and SK Hynix sit on top of. Watch foreign flows and market breadth, not the index number. When a benchmark is really two stocks in a trench coat, the headline is the least informative thing about it.
Get early access to Orbit
Orbit is Luna3.ai’s AI-augmented research engine. 12 algorithmic signals + a gradient-boosted ML model + an agentic LLM that reads each top pick’s filings and writes a daily thesis with conviction score and catalyst proximity. Three regimes, three playbooks — growth in expansion, defensives in late-cycle, recovery plays at panic bottoms. The 3 in Luna3.ai.
No spam. Unsubscribe any time.
No comments yet. Be the first to share your thoughts!