1
Lighter Than the Last Three Lows
Versus the 2022 tightening low: -16%; versus the 2025 tariff shock: -10%; versus the 2024 martial-law low: -7%. The index is higher than then, but debt is lower.
What Was Missing Was a Buyer
Reading markets as we view a forest from the sky
Jul 28, 2026 (Tue) 12:00 · Intraday · Seoul
A sudden gust blows.
Even in the gust, we watch the world calmly.
KOSDAQ set a new 52-week low. But what collapsed today was not debt — debt had already been quietly leaving for four months. What was missing today was a buyer.
Today's descent — the same observations as the intraday chart in section 01 below
01
KOSDAQ Today
As of 7/28 (Tue) 12:00 · 09:06 sell-side sidecar (KOSPI and KOSDAQ together) → 10:13 KOSPI circuit breaker (20-minute halt, 8th this year) · KOSPI at 12:00: −9.9%
KOSDAQ index
703.72
-7.99%
Intraday low 703.63 = new 52-week low
From the 52-week high
-42.7%
1,229 → 704
8pp deeper than KOSPI (−35% from its high)
Decliners
1,618
92 advancers
Of 1,710 stocks counted, 94.6% fell
Trading value
2.76tn
KOSPI KRW 15.9tn
The open was the high — no rebound attempt
From open to low, a day without a retracement
The opening 742.13 was also the intraday high — buyers never stepped forward. Note: Minute data are unavailable, so dots show only actual polling times while writing this report (not a continuous chart).
02
Who Sold? KOSDAQ Differs From KOSPI.
Today's intraday net buying by investor type · Unit: KRW 100 million
Buyers — retail investors
+1,211
Also absorbed +KRW 2.57tn on KOSPI — retail investors were the only buyers in both markets
Sellers — institutions + foreign investors
-1,196
Institutions −689 · Foreign investors −507 — institutions sold more on KOSDAQ
20-trading-day cumulative — foreign selling is concentrated in KOSPI
Foreign net selling on KOSDAQ is around 1/18 of KOSPI's. A foreign exodus does not explain KOSDAQ's fall.
View table
| Market | Retail investors | Foreign investors | Institutions |
|---|---|---|---|
| KOSDAQ | +11,459 | -10,011 | -2,843 |
| KOSPI | +154,498 | -181,942 | +21,867 |
Unit: KRW 100 million · 20 trading days (7/7–7/28) · Source: Naver Finance trading by investor type
Outflow
Foreign investors · Chronic
Net sellers on 16 of 20 days. Not KOSPI-style convulsions swinging by KRW 3tn a day, but a steady leak
Outflow
Institutions · Today's main sellers
−KRW 68.9bn today, more than foreign investors (−507). Institutions weigh more on KOSDAQ.
Inflow
Retail investors · +KRW 1.15tn over 20 days
Absorbing sales in both markets. But deposits fell −KRW 21.4tn over 20 days, reducing buying power.
Neutral
Program trading −KRW 54.3bn
Non-arbitrage −KRW 54.5bn. Unlike KOSPI (−KRW 1.73tn), basket-selling pressure is small on KOSDAQ.
03
The Epicenter: Semiconductor Equipment
There were three triggers, not one — the last to become known hurt the most
Trigger ①
New York semiconductors overnight
NVIDIA -4.99% · AMD -5.17% · SK hynix ADR −7.47%. Nasdaq fell for 4 consecutive trading days (−0.18%), but the Dow gained +0.51% — not the index, but a market that singled out semiconductors took the hit.
Trigger ②
CXMT surged on its first listing day
Yesterday's Shanghai STAR debut. Opening price: +452%; reported closing prices vary: +466~476%. The narrative that the proceeds would fund memory-wafer expansion was priced in alongside its No. 1 A-share market cap.
Trigger ③ · New today
China starts domestic production of immersion DUV lithography equipment
Reports of a breach in ASML's monopoly territory — deliveries to SMIC, Hua Hong, and CXMT planned this year. ASML shares −6%. Yet performance falls short of ASML's, at around 5 units a year, making this less an immediate volume shock than the pricing of a narrative of eventual self-sufficiency.
KOSDAQ's largest stocks by market cap: intraday returns
Semiconductor equipment worst at −11–13%; biotech and pharma held up
Sector returns — cyclicals and domestic defensives completely diverge
Whole-market sector classifications
Why did KOSDAQ equipment (−13%) fall more than KOSPI semiconductors (−9%)?
Equipment makers already faced a double bind. If domestic fabs feared oversupply and delayed capex, orders would fall; export controls prevent selling to China, the alternative customer. Today, it became a triple bind: if China produces DUV domestically, the problem shifts from "we cannot sell because of regulation" to "they will not buy even if we can sell". Finished-product makers face price competition; equipment makers face the possible disappearance of the customer itself. The same news therefore commands different multiples.
Yet the same article also contains counterevidence: 5 units a year with subpar performance. The gap between narrative and reality will later tell us whether this decline was excessive.
04
Yet Debt Had Already Left
10.5 years of KOSDAQ margin loans — KOSDAQ alone, extracted from market-separated fields in KOFIA's source data
KOSDAQ margin loans
6.93tn
10-year percentile: 45%
As of 7/24 · 10-year low KRW 3.31tn / high KRW 11.71tn
From this year's peak
-37.3%
KRW 11.05 → 6.93tn
Since the 4/29 peak, 86 days · Latest KRW 6.93tn
Leverage intensity
0.985
53% percentile
Margin loans ÷ 100 index points — 2019–20 levels
Margin loans / deposits
30.9%
Deposits KRW 105.6tn
Regime: cooling · both markets combined margin loans of KRW 32.67tn (KOSDAQ alone: 6.6%)
10.5 years of KOSDAQ margin loans (quarter-end, KRW tn)
Debt's absolute amount . The margin-financed positions built up in 2021 are being fully unwound.
KOSDAQ margin loans (quarter-end balances)
10-year high / low
Margin loans per 100 index points — debt's weight relative to market size
Margin loans divided by the index level (green, left axis). Higher means heavier debt. Intensity moves through both the numerator (debt) and denominator (index), so we overlay KOSDAQ index (gray, right axis) is overlaid. If the index collapses, intensity soars even with debt unchanged.
Leverage intensity — Left axis
KOSDAQ index — Right axis
Historical index-low dates
2021–25 margin-investing regime
Now: 0.99 — KRW 0.99tn of margin loans per 100 index points. It has fallen to the bottom of the margin-investing regime range (0.96–1.19), lighter than the last three lows (1.06–1.17), and almost equal to the pandemic low's level (0.95).
Note: Why low markers sit on peaks: indices can fall sharply in a day, while margin liquidation lags by several days, so intensity peaks on the day the index bottoms. In 2020, it reached 0.95 on 3/19, then fell to 0.60 after 12 days.
Note: Why low markers sit on peaks: indices can fall sharply in a day, while margin liquidation lags by several days, so intensity peaks on the day the index bottoms. In 2020, it reached 0.95 on 3/19, then fell to 0.60 after 12 days.
05
Alongside Past Lows
Each low shows leverage intensity on the day the index bottomed — just after the index fell sharply, before margin liquidation had progressed — capturing the heaviest moment. We compare like with like.
View table
| Low | Index | Margin loans | Leverage intensity |
|---|---|---|---|
| 2016-12-05 rangebound market | 575 | 3.75tn | 0.653 |
| 2019-08-06 trade dispute | 552 | 4.84tn | 0.878 |
| 2020-03-19 pandemic | 428 | 4.06tn | 0.948 |
| 2022-10-13 tightening bear | 652 | 7.62tn | 1.170 |
| 2024-12-09 martial law | 627 | 6.62tn | 1.055 |
| 2025-04-09 tariff shock | 643 | 7.00tn | 1.088 |
| 2026-07-28 now | 704 | 6.93tn | 0.985 |
2
Almost Equal to the Pandemic Low
Intensity in the March 2020 panic was 0.948; now it is 0.985 — a 4% difference. Virtually all leverage built up since 2021 has been reversed.
06
How Far Has Deleveraging Gone?
Compared with past peak-to-trough cycles · This episode has already progressed as far as the historical average
Margin-loan deleveraging by cycle
This episode (−37.5%) has compressed the entire 2021-22 tightening cycle (−36.9%) into 86 days Note: Here alone, the window low of KRW 6.91tn (−37.5%). The −37.3% tile in section 04 uses the latest KRW 6.93tn — the same episode measured at a different time
View table
| Cycle | Peak | Low | Decline | Duration |
|---|---|---|---|---|
| 2018 trade dispute | 6.52 | 4.32 | -33.8% | 272 days |
| 2020 pandemic | 5.91 | 3.31 | -44.0% | 30 days |
| 2021-22 tightening | 11.71 | 7.39 | -36.9% | 403 days |
| 2024 second half | 9.14 | 6.24 | -31.7% | 150 days |
| 2025 tariffs | 7.85 | 6.82 | -13.2% | 45 days |
| 2026 current | 11.05 | 6.91 | -37.5% | 86 days |
Still — we cannot call a bottom
Historically, margin loans fell for several more days after the index bottomed. Ordinary corrections ended with another 3–6% decline (to KRW 6.5–6.7tn): −3.1% in 2022, −5.7% in 2024, −2.5% in 2025. But the 2020 pandemic panic took them down −18.4%. We do not yet know which case this is.
07
So What Should We Watch?
Three findings from today's data
1
Forced-Liquidation Risk Is in KOSPI, Not KOSDAQ
Over the same period, KOSPI margin loans actually increased — KRW 22.28tn in March → KRW 25.75tn in July (+15.6%). Total margin loans (KRW 32.6tn → KRW 32.67tn) stayed flat; the mix shifted from KOSDAQ to KOSPI. In effect, investors borrowed to buy Samsung and SK hynix, which fell −9–11% today. KOSDAQ has already cleared 37%; pressure is building on the other side.
2
KOSDAQ's Problem Is the Void, Not Debt
Leverage is lighter, but the buyers have disappeared. Foreign flows are chronically negative, institutions are net sellers, and retail deposits fell −KRW 21.4tn over 20 days. Thin order books let the same selling push the index deeper. That is why its decline is 10pp deeper than KOSPI's and it reached a 52-week low first.
3
Some Numbers Have Not Appeared Yet
Margin-loan data run only through 7/24. Today's sharp falls of −8.0% (KOSDAQ) and −9.9% (KOSPI) are not reflected yet. Actual forced liquidations appear 2–3 business days later. How much KOSPI margin loans shrink in the next update is the real test of this episode.
08
This Week Will Judge
Within three days, we will know whether today's decline was narrative or reality
7/29 (Wed)
SK hynix earnings The company hit by −12.9% today. HBM4 progress and capex guidance will provide the first physical rebuttal or confirmation of trigger ③ (China's self-sufficiency narrative). Cutting capex would be worse news for KOSDAQ equipment makers than today.
7/30 (Thu)
Samsung Electronics earnings A second answer to the same question. Since 6/10, DDR5 spot prices have risen +14% without interruption. The key is how much commodity DRAM price commentary can absorb concerns over CXMT competition.
7/28–29
FOMC
A separate axis from today's sharp fall. With oil down −12% since 7/24, a hold with a hawkish tone is the baseline. But a hike signal under Chair Warsh would add another rate axis to today's decline.
7/30–31
BOJ
USD/JPY in the JPY 163 range, near the intervention zone (164–165). The yen's direction affects the won and Asian flows broadly.
7/30–31
Margin loans and investor deposits update The number discussed in 07-③. The first data reflecting today's −8%/−9.9% will show how much KOSPI's KRW 25.75tn of margin loans is cut — the real test of this episode.
Boboo's Rules for Watching
What moved today was price; what arrives this week is physical evidence. Trigger ③ (Chinese DUV) came with the physical caveat 5 units a year with subpar performance. Prices normally reflect narratives first, but a reversal comes if reality does not catch up with the story. Gusts change direction, not terrain. This week, we can find out what the terrain is.