1.2 Million Korean Investors Got Margin Calls on Leveraged ETFs
Wednesday, July 299 Min Read
Korean retail investors rode AI memory demand with leverage and were liquidated en masse in July. The finance minister apologised in parliament.
₩56.3 Trillion
Citi's Estimate of Leveraged ETF Losses
1.2 Million
Retail Accounts Hit by Margin Calls
~320 – 360k
Accounts Liquidated Outright
27%
Kospi's Drawdown From Its June High
In This Piece6 Sections
In mid-July, more than a million brokerage accounts in South Korea received the same message: a margin call.
According to figures circulated by Goldman Sachs's trading desk, over 1.2 million leveraged retail accounts had been hit by July 13. Between 320,000 and 360,000 were fully liquidated by their brokers.
South Korea's working-age population is roughly 35.7 million. That is about one in every thirty.
This was not an equity correction. It was a hole in the household balance sheet.
Leverage on a Single Stock: How the Product Was Born
At the centre of this sits a new product: the single-stock leveraged exchange-traded fund.
An ordinary ETF tracks a basket; a single-stock leveraged ETF multiplies the daily move of one company — here Samsung Electronics and SK Hynix. If the stock rises 3% intraday the fund rises roughly 6%, and if it falls 3% it falls roughly 6%.
Korea approved these products just as its two memory makers were running the sharpest rally in their history. AI data-centre demand had nearly tripled SK Hynix and doubled Samsung Electronics over the year. Riding a rising stock with borrowed money looked, in hindsight, flawless.
Money poured in. Total leveraged ETF assets peaked at $52.5 billion on July 22.
The problem was that most investors bought these funds not with their own cash but with margin loans. HSBC's equity strategy team put roughly $23 billion of margin lending behind retail participation.
The Arithmetic of Leverage: Two Separate Holes
Two mechanisms are at work here, and they stack.
The first is the fund's own internal decay. The second is the forced-selling threshold that comes with buying that fund on borrowed money.
The second hole works faster. When you buy a leveraged fund inside a margin account, your true exposure is the fund's leverage multiplied by the loan.
Someone buying a 2x fund with 50% borrowed money is effectively running 4x against the stock. At that level, a routine 10-12% pullback is enough to trigger a margin call.
When the call arrives you do not decide when to sell; your broker does, that day, at that moment. You can be liquidated while being right, because the time you need is longer than the time you can hold.
The Chain of Forced Selling
- 01Price fallsLeveraged fund falls twice as far
- 02Margin threshold breachedBroker issues the call
- 03Cash not postedPosition is closed out
- 04Selling pressure buildsPrice falls another leg
The last link feeds the first: forced sales push the price down, and the lower price pushes fresh accounts below the threshold. That was the machine running in Korea in mid-July.
July: The Month the Chain Snapped
Timeline
- JuneThe Kospi peaks. SK Hynix has nearly tripled on the year, Samsung Electronics has doubled.
- July 13Per Goldman Sachs desk data, more than 1.2 million leveraged accounts have taken margin calls; 320,000 to 360,000 are fully liquidated.
- July 16Samsung Electronics falls 8.8%, SK Hynix roughly 12%.
- July 22Leveraged ETF assets peak at $52.5 billion.
- July 24Margin loan balances top out at ₩38.6 trillion.
- July 27Balances fall to ₩32.7 trillion — the gap is largely forced selling.
- July 29The Kospi drops as much as 12.6% intraday. The finance minister and the FSC chairman apologise in parliament.
Where the losses concentrated tells you this was not a product story but a single-trade story.
The total is ₩56.3 trillion, about $38.7 billion. The product itself shrank just as fast.
Buying the Fall
The detail that deserves most attention: while assets fell from $52.5 billion to $19 billion, a further $6.2 billion of new money went in.
Buying continued all the way down. Retail investors' average purchase price in SK Hynix sits at ₩2.28 million, against ₩1.444 million on July 29 — an average loss of roughly 31.6%.
Averaging down is defensible in an unleveraged portfolio. In a leveraged product the same reflex extends decay and shortens the distance to the margin threshold.
Two different games, one identical instinct.
Leverage left a mark on the US side too: KORU, the New York-listed fund that triples the daily move of Korean equities, fell roughly 70% from its June 1 high. Its holders needed neither a Korean address nor a margin account — at three times leverage, volatility eats capital roughly nine times faster than at one.
Wherever the product trades, the arithmetic is the same.
The Other Reading
| Blaming the Product | Blaming the Choice |
|---|---|
| A single-stock leveraged ETF should not be sold to retail investors | The product was clearly labelled; using leverage was a decision |
| Approval came at the most extreme moment for memory stocks | The same instruments have traded in other markets for years |
| Margin loans were allowed to stack on top of fund leverage | Margin rules were published in advance and did not change |
Both columns hold together. Korea's political outcome favoured the first.
On July 29 Finance Minister Koo Yun-cheol told parliament he was "sorry for introducing the product without careful consideration."
Lee Eog-weon, chairman of the Financial Services Commission, said the regulator felt sorry that it had "fallen short in properly meeting the public's expectations."
What Is Left
Three regulatory steps followed: new single-stock leveraged ETF listings were suspended, marketing of the products was banned, and the minimum cash deposit requirement was tripled to ₩30 million, roughly $20,000.
Those measures govern what happens next. They do not reopen 320,000 closed accounts.
The memory demand story did not break. The AI-driven demand that lifted SK Hynix and Samsung is still there.
What broke were positions carried with more debt than they could bear. Risk management is the name for exactly that distinction.
This article draws on estimates circulated in Citi and Goldman Sachs client notes, figures published by HSBC's strategy team, official statements reported by Seoul Economic Daily and MarketScreener, and CNBC's coverage. Loss estimates reflect each institution's own methodology and vary between sources; the liquidation figure of 320,000 to 360,000 accounts is a broker estimate.
Correction (September 13, 2026): prose tightened for rhythm; figures and sources unchanged.
Companies in This Piece
Since The Event