The gauge said Korea was safe

I sold early.

It was a good stock. I got nervous and took my profit. Then I watched it keep climbing without me… and every green day felt like money somebody took out of my pocket.

So I understand what happened in Seoul this spring.

South Korea's stock market went on a tear. The world wanted memory chips for AI. Samsung Electronics and SK hynix couldn't make them fast enough.

Regular investors didn't want to miss it. So they borrowed.

A margin loan is money you borrow from your broker to buy more stock. Korea's margin book started the year at 27.4 trillion won. By June 22 it reached about 38.5 trillion won. That was a record.

The KOSPI set a record close that day too, above 9,100.

Then it broke.

From June 22 to July 30, the KOSPI fell 39%. On July 28 alone it dropped 10.84%.

Brokers sent margin calls. Investors who couldn't pay had their shares sold for them. Each forced sale pushed prices lower… and that set off the next margin call.

But the strangest part came from the gauge everybody used.

Analysts judge margin risk as a share of the whole market. Right before the selloff, Korea's margin loans came to about 0.6% of total market value.

Less than one percent. By that yardstick, Korea looked safe.

It wasn't. The average missed where the money went.

That's what worries me now.

The KOSPI closed back above 7,000 this week. Korea's margin book rose nine sessions in a row into late August, back to about 33.3 trillion won. Borrowing by investors in their 20s and 30s grew 2.4 times in six months.

The crash didn't cure the habit. It put stocks on sale.

And Korea isn't alone. Margin books in China, Japan and India are growing too.

I have no idea when the next one breaks. But I know which number I've stopped trusting.

Leverage looks harmless when you spread it across a whole market. The trouble starts when it all sits in one place.

In Korea, it sat in Samsung Electronics and SK hynix. Margin loans on those two stocks grew 3.1-fold this year through late May.

On May 27, Korea started listing leveraged ETFs tied to single companies. Now a retail investor could borrow and then bet on a fund that was itself leveraged.

When the selling came, those two stocks fell hardest. Samsung dropped 43%. SK hynix dropped 55%.

Marcellus Investment Managers counted 1.2 million leveraged accounts hit with margin calls. About 360,000 were liquidated.

The book is growing again. So are the chip stocks. It's the same setup.

And here's where it spreads.

India's margin trading book hit a record ₹1.49 trillion on September 15. A year earlier it stood near ₹99,000 crore. That's about 50% growth in twelve months.

The market went the other way. The Nifty is down about 10.7% this year. The Sensex is down about 12.8%.

Margin books usually shrink when prices fall. Borrowers get scared, or they get sold out. India's book has grown six months in a row.

Borrowing is going up while the market goes down. That's backwards.

By one estimate, more than half of that borrowed money sits in small and mid-sized companies. Those shares trade thinly. When sellers rush for the exit, buyers aren't there.

The Reserve Bank of India already stopped people from using loans against their shares to buy more shares or IPOs. The margin book kept growing anyway.

Meanwhile, the biggest margin book in Asia belongs to China.

On May 11, China's margin financing and short-selling balance hit a record 2.83 trillion yuan. That's about $416 billion. It topped the peak from the 2015 boom.

Regulators saw it building. In January they raised the cash requirement on new margin trades to 100%, up from 80%. The balance kept climbing.

The defense is the ratio. Caixin puts margin at about 2.5% of tradable market value, versus 4.7% in 2015.

I'd find that more comforting if Korea hadn't just crashed on a ratio a fraction that size.

In China and Japan, the borrowed money leans toward big tech and chip stocks too. Japan's margin balance has grown past $35 billion, from $18 billion in early 2020.

Same setup. Different country.

Three more things worth keeping track of…

Old AI chips are getting more expensive. The hourly rate to rent an Nvidia H100 rose 21.9% in a month, to $3.28 from $2.69, according to the computing exchange Ornn. That's interesting. Older chips usually get cheaper when newer ones arrive. Korea's Vessl AI raised its own H100 rate 24.7% on September 1.

SoftBank keeps borrowing against its Arm shares. That loan started at $8.5 billion in 2023. This month SoftBank raised it to $25 billion. It's also selling more than $11 billion of junk bonds to fund its OpenAI commitment. The yield on its 2031 dollar bond rose to 8.2% this month, from 6.7% in January.

Europe's gas storage was about 68% full in mid-September. A year earlier it was about 81%. The five-year average for that date is about 84%. Dutch TTF gas futures crossed €82 per megawatt-hour. Heating season is weeks away. We'll see.

In June 2015, China had a number everyone watched.

It was the official margin balance. On June 18 it hit 2.27 trillion yuan. That was a record.

Regulators tracked it. Brokers reported it. Analysts put it in their charts.

But a second margin book ran alongside it.

Trust companies and online lenders were lending to traders outside the official system. Much of it ran through software called HOMS, built by Hundsun Technologies in Hangzhou.

HOMS let one brokerage account split into many smaller ones. A trader could borrow heavily without ever registering with China's clearing house.

These shadow accounts didn't follow the forced-selling line that governed regular margin accounts. Regulators couldn't trace who owned them. One widely cited estimate put the shadow book near 1.5 trillion yuan.

The Shanghai Composite peaked on June 12 at 5,178. Then regulators moved against the shadow accounts. The market lost about 30% in roughly a month.

By mid-July, 1,476 of the 2,808 companies listed in Shanghai and Shenzhen had suspended trading in their own shares. More than half the market simply stopped.

The official number was accurate. It just wasn't the whole number.

Economists writing for the NBER later traced the crash to those shadow-financed accounts.

Every margin figure I look at today is an official one. Exchanges publish them every day.

I have no idea what sits outside them. Neither does anyone else.

The number you see is never the full bill. We'll see.