Korea's Semiconductor Index Caps SK hynix at 20%. Here Is the Math.
Tomorrow, September 10, near the close, ETFs tracking the KRX Semiconductor Index will sell an estimated ₩1.24 trillion of SK hynix and ₩207 billion of Samsung Electronics. Not because anything changed at either company, but because Korea Exchange caps any single stock at 20% of the index, and SK hynix had drifted to 36.75%. The headline number is large. The number that matters for tomorrow's close is smaller and sits somewhere else: Hanmi Semiconductor's estimated buy is 2.74 times its average daily turnover.
Everything here is from public sources: Korea Exchange announcements, Mirae Asset Securities estimates as reported by the Korean financial press, and fund documentation. I flag every derived number.
KEY TAKEAWAYS
1. The KRX Semiconductor Index enforces a 20% single-stock cap at its annual September reset. As of September 4, SK hynix was 36.75% and Samsung 22.78%. Mirae Asset Securities estimates ₩1.45 trillion of combined selling in the two names at the September 10 close.
2. Scaled to liquidity, the big tickets are small: 0.19× 20-day average turnover for SK hynix and 0.04× for Samsung. The buy side is where the stress is. Hanmi Semiconductor's estimated inflow is 2.74× its daily turnover.
3. Across all 29 KRX indices being reset, ₩1.8 trillion of the ₩2.2 trillion in expected turnover comes from weight changes, not from the 18 semiconductor names being added or dropped.
What is actually happening on September 10
Korea Exchange publishes 17 KRX sector indices. Once a year, in September, it rebuilds them: constituents are re-selected on July-end data and weights are reset. On September 3 and 4 the exchange announced the results for 29 indices, including the sector set and KOSDAQ TOP10. The new compositions take effect September 11, so index-tracking ETFs trade at the September 10 close. Mirae Asset counts 33 Korean ETFs that directly track these 29 indices, with ₩11.7 trillion in assets as of September 4.
The semiconductor index rules, in one table:
| Rule | Detail |
|---|---|
| Reset frequency | Annual, every September (per Samsung Asset Management's KODEX Semiconductor fund documentation) |
| Constituent screen | Stocks in the sector that fall within the top 95% of cumulative market cap and the top 90% of average daily turnover, measured at July-end (exchange methodology, summarized) |
| Weighting | Float-adjusted market cap |
| Single-stock cap | 20%. Excess weight is cut at the reset and redistributed pro rata to the remaining constituents |
| Timing | ETF trades near the September 10 close; new index effective September 11 |
This year's semiconductor changes: two additions (TES, TSE) and sixteen deletions (Nextin, Gaonchips, LX Semicon, SFA Semicon, Duksan Neolux, Philoptics, YC, Hana Materials, KC Tech, S&S Tech, HD Hyundai Energy Solutions, KoMiCo, RFHIC, Taesung, Techwing, Koh Young). Simmtech joins the separate KRX IT index.
One sourcing note. Tables of "estimated buy/sell amounts" circulate in Korea with "Source: KRX" attached. The exchange publishes constituents and weights. It does not publish trade estimates. Those come from brokers multiplying tracking-ETF assets by weight changes, which is why the numbers differ from desk to desk. The charts in this post use only the Mirae Asset figures that were reported in the press; the range column below is a secondary compilation I could not trace to original reports.
| Stock | Change | Range of broker estimates (secondary) | Mirae Asset (Sep 4) |
|---|---|---|---|
| SK hynix | Cut to 20%, sell | ₩1.13T to ₩1.45T | ₩1,244B |
| Samsung Electronics | Cut to 20%, sell | ₩190B to ₩240B | ₩207B |
| Hanmi Semiconductor | Weight up, buy | ₩209B to ₩300B | ₩306B |
| Jusung Engineering | Weight up, buy | ₩121B to ₩180B | ₩180B |
| TES | New addition, buy | about ₩140B | — |
| Wonik IPS | Weight up, buy | about ₩82B | — |
| EO Technics | Weight up, buy | about ₩81B | — |
The arithmetic of a 20% cap
SK hynix loses 16.75 percentage points (36.75% to 20%). Samsung loses 2.78 points (22.78% to 20%). The 19.53 points freed up are spread across the rest of the index in proportion to float-adjusted market cap. That is where the ₩306 billion Hanmi buy and the ₩180 billion Jusung buy come from.
The estimates also let you back out the asset base. ₩1,244 billion divided by 16.75 points is roughly ₩7.4 trillion. ₩207 billion divided by 2.78 points is also roughly ₩7.4 trillion. The two agree, so Mirae's model appears to assume about ₩7 trillion of assets tracking the semiconductor index. On the same base, Hanmi's ₩306 billion works out to a weight gain of about 4 points. Both figures are my derivations from reported estimates, not published by the exchange or any fund manager.
Why is the overshoot so large this year? Because the two megacaps outran everything else in the index over the past twelve months, and kept going after the measurement date. From August 7 to September 7, SK hynix rose 19.26% and Samsung 17.14% (closing at ₩1,783,000 and ₩270,000 on September 7). Gains after September 4 mean the actual won amount traded on the 10th could exceed the estimate.
Size the trade against liquidity, not against the headline
₩1.24 trillion is a lot of SK hynix. But Mirae's own report scales each trade against the stock's 20-day average turnover, and on that measure SK hynix is 0.19× and Samsung is 0.04×. That is a fifth of a normal day and a twenty-fifth of a normal day, respectively.
On the buy side, Hanmi Semiconductor is 2.74× and Jusung is 0.90×. Hanmi is being asked to absorb nearly three normal days of volume, in one direction, in the closing auction. Mirae's analyst flagged exactly this: stocks with a high trade-to-turnover ratio can see widened volatility right before the close. The market appears to have started positioning already; on September 7, Hanmi rose 4.57% and Jusung 4.99%.
For scale on the sell side: the ₩1.45 trillion of combined selling is about a third of the ₩4.09 trillion that foreign and institutional investors net bought in the two stocks on September 7 alone (my calculation from the reported figures). If that buying persists, one closing-auction print is not the kind of size that changes direction.
The money is in weight changes, not in the 18 names
Two additions and sixteen deletions make for a long list, but Mirae estimates that of the ₩2.2 trillion in expected turnover across all 29 indices, only ₩0.4 trillion comes from additions and deletions. ₩1.8 trillion is weight adjustment among existing constituents, and the ₩1.45 trillion of semiconductor megacap selling is two-thirds of the total by itself.
The sixteen deletions and the cap share a root cause. The constituent screen keeps stocks inside the top 95% of cumulative sector market cap. As the two largest names grow, that 95% line rises and small caps fall outside it. Concentration produces deletions. (This is my reading of the methodology, not an exchange statement.)
Concentration is showing up in the exchange's other indices too. According to a Money Today report on September 9, Korea Exchange has notified derivatives dealers that the KRX 300 will be reclassified as a narrow-based index under US CFTC rules from September 15, because its largest constituent has grown too big. The CFTC test is met if any single stock exceeds 30% or the top five exceed 60%; Samsung was about 28% of the KRX 300 as of September 8. KOSPI 200 and the Korea Value-up Index were reclassified earlier this year. Separately, the exchange is preparing a Korean industry classification system, KRICS, for November, which could redraw the sector indices again.
What I actually watch
| Checkpoint | Why |
|---|---|
| September 10, 15:20 to 15:30 KST closing auction, Hanmi and Jusung | Highest trade-to-turnover ratios. The first place price should move, and the first place a next-day reversal would show. |
| Who takes the other side of the megacap selling | Foreigners net bought ₩1.35T of SK hynix and ₩871B of Samsung on September 7. Whether that continues on the 10th decides if the mechanical sell is absorbed or visible. |
| KRX 300 futures after September 15 | Narrow-based status adds SEC oversight for US participants. Watch foreign volume and liquidity in the futures. |
| KRICS rollout in November | If sector membership rules change, this arithmetic has to be redone. |
Value chain read-through
| Group | This reset | What to watch |
|---|---|---|
| Samsung Electronics, SK hynix | Capped, sell | Large in won, small in liquidity terms (0.04× to 0.19×). One-day mechanical flow. |
| Hanmi Semiconductor, Jusung Engineering, Wonik IPS, EO Technics | Weight up, buy | Trade-to-turnover ratio, closing volatility, next-day give-back. |
| TES, TSE | New additions, buy | Price behavior after the first trading day in the index (September 11). |
| 16 deletions | Removed, sell | Closing-print impact in thin names and how fast it recovers. |
Risks to this view
- Estimate error. Funds differ between full and partial replication, and prices have moved since the September 4 measurement date. Actual amounts will differ.
- Pre-positioning. The September 7 gains in Hanmi and Jusung show buying ahead of the event. Event-day price action can run the other way.
- This is flow, not fundamentals. The rebalance is a one-day mechanical event. From September 11 the drivers are earnings and memory pricing again.
- Source hierarchy. The trade sizes are broker estimates, not exchange data, and the range column is a secondary compilation.
Closing
The reset happens once a year, but as long as the 20% cap exists, any stock that outperforms its sector meets the same mechanical selling every September. This year the cut was 16.75 points. For reading tomorrow's close, 2.74× daily turnover is a more useful number than ₩1.4 trillion. If KRICS reshuffles the sector indices in November, I will run this arithmetic again.
Sources: Korea Exchange, regular index review announcements (Sep 3–4, 2026); Mirae Asset Securities estimates (analyst Yoon Jae-hong, as of Sep 4) via Herald Business, eToday and SR Times (Sep 8, 2026); Money Today on KRX 300 narrow-based reclassification (Sep 9, 2026); Edaily and Financial News on KRICS (Sep 7–8, 2026); Samsung Asset Management KODEX Semiconductor fund documentation;
Comments
Post a Comment