Korean retail borrowed a record ₩38.6 trillion (~$28B) to chase the top — and on July 13 it snapped, hard.
- The SetupAs the KOSPI ripped to a record 9,100, retail investors' margin debt — known locally as "bittu" (빚투), literally "debt-investing" — hit an all-time high of about ₩38 trillion (roughly $27B), up more than ₩10T (~$7B) from ~₩27T (~$19B) at the end of 2025. Counting stock-collateralized loans too, total retail leverage is near ₩60T (~$43B).
- The ConcentrationThe borrowed money isn't spread out — it's piled into two chip names. Margin tied to Samsung Electronics jumped from ₩1.65T (~$1.2B) to ₩4.76T (~$3.4B) in six months; SK Hynix went from ₩0.88T (~$0.6B) to ₩4.33T (~$3.1B). Together ~₩9T (~$6.4B), roughly a quarter of all margin debt, sits in two correlated semiconductor stocks.
- The RiskThis is the danger, not the story. Margin loans cost 7–9% and auto-liquidate if collateral drops. It broke three times in 2026 — a −12% plunge in March, a chip rout in June, and the big one on July 13 (−9%, circuit breaker) that is force-selling leveraged accounts right now. Brokerages had already hit their lending caps — the shock absorber was gone.
Imagine borrowing money from your broker to buy more stock than your cash allows. You put up ₩10M (~$7,000), the broker lends you another ₩10M, and now you control ₩20M (~$14,000) of shares. If the stock rises, your gains are doubled. If it falls, your losses are doubled — and here's the trap: if your holdings drop below a set line, the broker doesn't call to ask. It sells your shares automatically to get its loan back, whether you like it or not. In Korea that forced sale is called 반대매매 (bandae-maemae).
That's "bittu" — 빚 (debt) + 투자 (investing). It feels great on the way up, which is exactly why ₩38 trillion (about $27B) of it has piled up near a record high. The problem is what it does on the way down: everyone's auto-sells fire at the same time, pushing prices lower, which trips more auto-sells — a chain reaction. Leverage adds fuel to the rally and gasoline to the crash.
| Metric | Figure |
|---|---|
| Retail margin loans (신용융자) | ~₩38T record ≈ $27B — up from ~₩27T (~$19B) end-2025 |
| Total retail leverage (incl. stock-backed loans) | ~₩60T ≈ $43B |
| KOSPI level | Record 9,385.59 (Jun 19 intraday); closed 6,806.93 on Jul 13 — −27% in 3 weeks |
| Margin as % of market cap | Under ~1% overall — but heavily concentrated |
| Margin interest rate | 7–9% annually |
| Samsung Electronics margin | ₩4.76T ≈ $3.4B (from ₩1.65T / ~$1.2B in 6 mo) |
| SK Hynix margin | ₩4.33T ≈ $3.1B (from ₩0.88T / ~$0.6B in 6 mo) |
| Brokerage lending capacity | Caps hit — several firms halted new margin buys |
Here it is in one breath: Korean traders were borrowing more money to bet on stocks than at any point in 20 years — and crammed nearly a quarter of it into just two chip stocks, right at the record high. When those two cracked in July, the borrowed money got sold automatically: 1.2M accounts margin-called, 320,000+ wiped out entirely, and the market fell 27% in three weeks. Leverage built the rally, then blew it up.
The single biggest risk isn't the total — it's the concentration. A quarter of all borrowed money is in two stocks that move together on the same AI-memory narrative. If chips roll over, the margin calls hit both at once.
This isn't a hypothetical. The leverage was stress-tested three times this year — and the third was the big one:
| Event | What happened |
|---|---|
| March 4 — Iran-crisis plunge | KOSPI fell ~12% in a day (6,300 → low-5,000s). Circuit breaker + sidecar tripped. Korea Investment & NH halted new margin buys as limits maxed; forced-liquidation ratio jumped from under 1% to 6.5%; ₩100B+ (~$70M) auto-sold in two days |
| June 23 — AI/chip selloff | Global AI-selloff dragged KOSPI down sharply from its record; Samsung and SK Hynix fell ~10%; circuit breaker triggered again; forced stock sales hit their highest in nearly 3 years |
| July 13 — the big one | KOSPI −8.95% in a day; circuit breaker + sidecar. SK Hynix −15.37% (worst day ever), Samsung −10.7%. KOSPI fell below 6,800 by Jul 14. July forced liquidations already ₩344.2B (~$230M); the forced-sell rate hit 10.5% of receivables (Jun 9) and 10.2% (Jul 9) — six of the year's top-10 forced-sell days came after single-stock leveraged ETFs listed May 27 |
The first two times the market recovered — which is exactly what emboldened the next, bigger wave of borrowing. The third started from the highest base of leverage yet, with brokerages already at their caps: that's why July's unwind has been the most violent, and why every single-stock leverage product is now underwater.
Over 300,000 investors just got wiped out chasing this rally on borrowed money, and it triggered a 27% crash. That, to me, is exactly when you want to be buying. Here are the three Korean assets I'm buying — ordered lowest risk to highest.
- Record leverage into a record high — now reversing. ₩38.6T (~$28B) of borrowed money chased the KOSPI to a record 9,385. The gains were borrowed and so is the downside: the deleveraging cascade that followed is the forced-selling now underway.
- The concentration makes it fragile. ~24% of all margin debt sits in Samsung and SK Hynix. This isn't a diversified market bet — it's a leveraged, one-way wager on AI memory. A chip pullback force-sells both at once.
- The safety net is nearly used up. Brokerages have hit lending caps and started halting margin buys; margin costs 7–9% and auto-liquidates. Two forced-selling cascades already fired in 2026. The mechanism to unwind this violently is fully in place.
Want to know how Johnny is hedging Korea/chip exposure into this?
Join the Discord to find out! →Korea Financial Investment Association (margin balance data) · Seoul Economic Daily (₩38T record; 20-year high; ₩36T regulator warning; March −12% margin halt; margin-call surge) · The Korea Herald & Korea Times (Samsung/SK Hynix margin concentration; brokerage lending caps; forced sales 3-yr high) · Nikkei Asia · Bloomberg / CNN (June chip selloff & circuit breaker) · Korea Capital Market Institute (borrowing limits). Figures as of June–July 2026; won/USD conversions approximate at ~₩1,400/$ (sources cited a weaker ~₩1,530/$ in places, so USD figures are indicative).