When Two Companies Are Half Your Index — What Korea Just Demonstrated
 |
Yesterday a national stock index fell 12.63 percent in a single session. The trigger was not a war, a default, or a central bank. It was a report that a factory in another country had begun mass-producing a machine.
Korea's KOSPI closed at 5,262.77 on July 29, 2026, after news that China had started mass production of domestic deep-ultraviolet chipmaking equipment. The index is down roughly 34 percent from its June peak of 9,114, and it fell about 23 percent across July alone — the worst month in its recorded history. Seven circuit breakers fired during those weeks. Somewhere around 250 trillion won of market value stopped existing.
I live here. I do not own a single Korean stock, which is an odd place to watch this from — close enough to hear it and far enough to look at the mechanism instead of my own balance.
And the mechanism is not really about Korea.
One Piece of News Should Not Move a Whole Country
Here is the part worth slowing down on. A single equipment announcement, in a single foreign industry, repriced an entire nation's benchmark by double digits in one day. That should feel disproportionate. It only makes sense once you look at what the index actually holds.
As of May 2026, Samsung Electronics was about 26.0 percent of the KOSPI's market capitalization and SK Hynix about 20.9 percent. Together, roughly 47 percent — call it half the index in two names. In the narrower KOSPI 200, the pair crossed 50 percent for the first time that same month.
Two years earlier, in 2024, those two companies were closer to 40 percent combined. The AI memory cycle did the rest.
So the KOSPI is not a bet on Korea. It is a bet on two memory manufacturers, wearing a country's name. When someone credibly threatens the moat around those two businesses, the "national index" does what a two-stock portfolio does.
That is not a criticism of Korea, and it is not a forecast about semiconductors. It is a description of arithmetic that was visible on any fact sheet, months before yesterday.
This Happens by Design, Not by Accident
 |
The next question is how an index gets here, and the answer is uncomfortable: it gets here on purpose.
Almost every major index — the KOSPI, the S&P 500, the vast majority of the funds tracking them — is weighted by market capitalization. The bigger a company becomes, the more of the index it is. That single rule has a consequence people rarely say out loud.
A cap-weighted index automatically buys more of whatever just went up.
Not because anyone decided to. Because the weight is the price times the share count. When Samsung and SK Hynix tripled on AI memory demand, no committee voted to make them half the KOSPI. They became half the KOSPI by rising, and every fund tracking the index had to follow them up to stay tracking it.
So a cap-weighted index is a momentum machine with a diversification label on the box. In a broad market it looks and behaves like diversification. In a market where one industry wins for three years running, it quietly converts itself into a sector bet, and it does so without ever sending you a notice.
This is also why concentration tends to peak right before it hurts. The weight is highest exactly when the winners have run the furthest — which is the same moment the downside is largest. The mechanism that built the concentration is the mechanism that makes the unwind steep.
None of that makes cap-weighting wrong. It is cheap, it is transparent, and it has beaten most alternatives over long stretches. It simply is not the thing most people believe they bought.
The Part Most Coverage Skips
There is a second-order effect here that almost never appears in market write-ups, and it is the most interesting thing I learned this week.
Concentration can become a forced-selling trigger.
Goldman Sachs flagged that if the combined Samsung–Hynix weight rose by roughly one more percentage point, foreign institutional investors would be compelled to sell somewhere near two billion dollars of holdings. Not because they had turned bearish. Because United States diversification rules — the ones written into the Investment Company Act — cap how much of a regulated fund can sit in a small number of positions.
Read that again slowly. An index can become so top-heavy that some of its own largest buyers are legally barred from owning more of it, and are required to sell at whatever price the market offers on the day the threshold breaks.
That is a mechanical seller, arriving for reasons that have nothing to do with whether the companies are good. It is the same species of problem as a margin call, moved up to the institutional level. And it is knowable in advance, because the rules are published.
Around 2.3 trillion won of positions were forcibly liquidated during this drawdown, and the president stepped in with emergency measures. Forced sellers do not negotiate. They sell.
What I Wrote in May, and What I Missed
On May 22 I published a post about the semiconductor rally, and I want to be exact about what it did and did not say.
I did flag concentration risk. The post named it directly, alongside a comparison to 1999 and a warning about paying up for real demand. It closed on this: "Not whether the fire is real, but how much you are paying to stand near it."
I still think that was the right question. But I asked it about the wrong container.
I was looking at concentration inside the semiconductor supply chain — too much of the stack depending on the same demand. What actually broke was concentration inside an index that most Korean investors hold through pensions and funds they never read the composition of. The risk was not that the fire was overpriced. It was that half the room was made of the same wood.
Then on July 8 I wrote about SK Hynix's American listing, and noted that it opened U.S. access to an AI-memory play in an inherently cyclical industry. That was accurate and it was insufficient. Access is not the variable that hurt people. Weight was.
The Numbers a U.S. Reader Should Actually Check
None of this is a Korean curiosity. It is a more advanced case of a condition the American market also has.
As of July 2, 2026, the top ten companies made up about 36.4 percent of the S&P 500, and the top five alone about 25.75 percent. At the end of 2025 the top ten were closer to 40 percent. These are not Korea's numbers. Korea's two-stock half-index is a different order of magnitude.
But look at the direction rather than the level. A quarter of the S&P in five names is a concentration that has not been seen in roughly two decades. The index that most American savers treat as diversification itself has been quietly becoming a smaller bet.
Two indices, two continents, one arithmetic. When a handful of names carry the weight, an index stops behaving like an index and starts behaving like those names.
The Buyer and the Sellers
There is a detail from this same week that makes the point better than any chart does.
On July 28, 2026 — the day before the KOSPI fell 12.63 percent — Apple closed at an all-time high of $340.08, touched $342.89 intraday, and became the first American company to carry a five-trillion-dollar market capitalization.
Apple is also the company whose reported interest in Chinese memory suppliers helped start July's repricing in Korea. And it sits on the other side of the transaction. Memory is a cost line on Apple's income statement and a revenue line on Samsung's and SK Hynix's. Cheaper memory — or even the credible threat of cheaper memory — reads as relief in one set of accounts and as pressure in the other.
I am not claiming one clean cause for either move. Apple's year has its own drivers, and the analysts arguing about them are not arguing about Korean memory. The narrower point is harder to dispute: the same industry story reached two indices with opposite signs, because the two indices sit on opposite sides of the same invoice.
Now put the weights back in.
Apple alone is roughly seven percent of the S&P 500. Samsung and SK Hynix together are roughly forty-seven percent of the KOSPI. So an American index holder owned about seven percent of the company printing records. A Korean index holder owned about half of the companies absorbing the hit.
Neither of them picked that allocation. Cap-weighting picked it for both, quietly, over about two years.
Where This Reading Breaks Down
I want to mark the limits, because concentration is easy to turn into a scare story and that is not what the data supports.
Concentration is not the same as overvaluation. An index can be top-heavy and reasonably priced, or broad and expensive. They are separate measurements and people constantly collapse them into one.
Concentration is also not a timing signal. The Samsung–Hynix weight crossed uncomfortable thresholds many months before the drawdown, and anyone who sold on the concentration reading alone would have missed a very large run first. The KOSPI reached 9,114 while the concentration was already extreme.
And it is not an argument against index funds. A concentrated index still beat most people's stock picking. The point is narrower: an index fund is a promise about process, not a promise about diversification. Those got conflated somewhere along the way.
Finally, none of this predicts a recovery or the absence of one. Local coverage is skeptical of a V-shaped rebound given expected rate decisions and cash-flow questions at the large caps. I have no view worth publishing on that, and neither does anyone selling you one.
How to Hold This
 |
Four things that can be done on a calm afternoon, none of which require a market call.
Look up your own top-ten weight. Every major fund publishes it. If you own a broad index fund and cannot say roughly what share sits in the ten largest positions, that is the gap to close first — it takes about two minutes on the issuer's page.
Separate the two questions. "Is this expensive?" and "Is this concentrated?" have different answers and different remedies. Conflating them produces the wrong action in both directions.
Know where your forced sellers are. If your holding sits in a structure with leverage, margin, or a regulatory cap, someone else may control the timing of your exit. That is worth knowing before the week it matters, not during it.
Decide the weight before the headline. The single most useful thing about concentration data is that it is available early. Yesterday's 12.63 percent day was violent and unpredictable in its timing. The composition that made it possible had been published for months.
Lines to Watch
Not predictions — just the numbers that would tell you the picture has changed.
- The combined Samsung–Hynix weight in the KOSPI and KOSPI 200. This is the whole story in one figure, and it moves both ways.
- The top-ten and top-five share of the S&P 500, quarter by quarter. Direction matters more than any single reading.
- Forced-selling thresholds — regulatory diversification limits, and any disclosed institutional mandate caps. These are the mechanical sellers.
- Circuit breaker frequency. Seven in a month is a measure of how thin the two-way market has become.
- Whether the DUV equipment story turns into shipped volume or stays an announcement. Those are different facts with the same headline.
Closing
The thing that struck me, watching from a country whose benchmark just did this, is how little of it was hidden.
The weights were on the fact sheets. The diversification rules were in the statute. The 2024-to-2026 climb from forty percent to half an index happened in public, one quarterly rebalance at a time. Nobody needed a forecast to see that a two-name index would behave like two names.
What was missing was not information. It was the habit of reading the composition of the thing you already own.
That habit costs nothing, and it is available today, in whatever market you happen to be standing in.
KOSPI July 2026 figures — 5,262.77 close and 12.63% single-day decline on July 29, 2026; June peak of 9,114; roughly 23% monthly and 34% peak-to-trough declines; seven circuit breakers; approximately 250 trillion won of market value and 2.3 trillion won of forced liquidations — per Korea Herald, Seoul Economic Daily, and Al Jazeera coverage, July 2026. Samsung Electronics (26.04%) and SK Hynix (20.90%) KOSPI weights as of May 2026, and the KOSPI 200 crossing 50% combined, per Asia Business Daily; 2024 combined weight of approximately 40% per the same reporting. Goldman Sachs forced-outflow estimate and the U.S. Investment Company Act diversification constraint as reported July 2026. S&P 500 top-ten (36.40%) and top-five (25.75%) weights as of July 2, 2026, per State Street data; end-2025 top-ten figure of approximately 40%. Apple all-time-high close of $340.08 and intraday high of $342.89 on July 28, 2026, and the five-trillion-dollar market capitalization milestone that day, per MacDailyNews and 24/7 Wall St.; Apple's S&P 500 weight of approximately 7% per index composition reporting, early 2026. China DUV equipment mass-production reporting, July 2026. Index weights change with every rebalance — readers should verify current figures on the issuer or exchange page before acting.
Related Posts:
The 2026 Semiconductor Rally — Everyone But NVIDIA
SK Hynix's ADR Listing — The Split That Isn't a Split
How to Know If Your Position Is Too Big — The Sunday Night Test
