2つのチップが韓国市場の半分を占めていた。そして、それは暴落した。

Key Takeaways

• The amplification was the story: On June 23, 2026, South Korea’s benchmark KOSPI index closed down 9.99 percent while the US Nasdaq Composite fell 2.2 percent the same day. Korea fell roughly four and a half times harder.
• Two stocks, one market: Samsung Electronics and SK Hynix together crossed 50 percent of the entire KOSPI’s market value in late May. When they move, the whole index moves.
• Memory is the highest-beta bet on Artificial Intelligence (AI): Broadcom grew its AI chip revenue 143 percent and its stock still fell almost 13 percent. The market now demands acceleration, not just growth.
• The danger is concentration, not demand: AI memory demand is still climbing. The risk is that an entire national economy is now riding a single, violently cyclical product.

A Coronation, Then a Circuit Breaker

On June 22, 2026, SK Hynix overtook Samsung Electronics to become South Korea’s most valuable company for the first time. That same day, the KOSPI closed at a record high of 9,114.55. A company that builds the memory chips inside AI data centers had just been crowned king of an entire national market, at the exact top.

The crown lasted about 24 hours. The next session, June 23, the KOSPI closed at 8,203.84, down 9.99 percent, finishing at its low for the day with no bounce into the close. The drop was steep enough to trip a circuit breaker, the automatic trading halt exchanges use to slow a panic. SK Hynix fell about 12.5 percent and Samsung about 12.3 percent in a single day.

Here is the part that should make you stop scrolling. The trigger was not Korean. It was a global wobble in AI stocks driven by worries about heavy AI spending and the threat of Federal Reserve rate hikes, the same fear that knocked the US Nasdaq Composite down 2.2 percent that day. A 2 percent American tremor became a 10 percent Korean earthquake. The question is why.

The Concentration Math

The answer is brutal arithmetic. By late May 2026, the combined market capitalization of Samsung and SK Hynix had surpassed 50.44 percent of the entire KOSPI, against a total market value of roughly 6,728 trillion won. Two companies. Half the index. Analysts had already flagged the setup as a concentration risk that leaves the benchmark dangerously exposed to any slowdown in data center investment.

Think of an index fund as a basket meant to spread your risk across hundreds of companies. The KOSPI is no longer that basket. It is closer to a leveraged bet on one product: high-bandwidth memory (HBM), the stacked, ultra-fast memory that AI accelerators consume by the truckload. When you buy “Korea,” you are mostly buying two memory makers selling to one end market.

That is why the math ran the way it did. The same global shock hits every market, but it lands on a concentrated index with far more force:

$\text{Amplification} = \frac{-9.99\%\ (\text{KOSPI})}{-2.2\%\ (\text{Nasdaq})} \approx 4.5\times$

Index or stock (June 23, 2026)	One-day move
Nasdaq Composite (US)	-2.2%
KOSPI (South Korea)	-9.99%
Samsung Electronics	-12.3%
SK Hynix	-12.5%

Sources: Yahoo Finance price data and WSJ market coverage.

Why Memory Falls Hardest

Memory is the boom-and-bust corner of the chip industry. Logic chips like processors are designed and differentiated. Memory is closer to a commodity, sold by the bit, prone to violent swings as supply and demand lurch out of sync. Stack an AI gold rush on top of that cycle and you get a high-beta bet: when AI sentiment rises, memory rises more, and when it falls, memory falls more.

The clearest illustration came three weeks before the crash. On June 3, Broadcom reported quarterly revenue of $22,187 million, up 48 percent, with AI semiconductor revenue of $10.8 billion, up 143 percent year over year. It even guided next-quarter AI revenue to grow over 200 percent, to $16.0 billion. By any normal standard, that is a spectacular result.

The stock fell nearly 13 percent the next day. Not because growth disappeared, but because the outlook held steady instead of leaping higher, and a market priced for perfection treats merely-spectacular as a disappointment. This is the deceleration trap: once a sector is valued for acceleration, growing 143 percent is not enough if last quarter the implied promise was more. The most AI-levered names get punished first and hardest. On June 23, US memory maker Micron fell about 13 percent in the same selloff. Memory is where the leverage lives, on both sides of the Pacific.

The Nokia Warning

If you want to know where this road can lead, look at Finland.

A generation ago, Nokia was Finland. At its peak, Nokia handsets and related equipment accounted for 20 percent of Finnish exports, and the company was the dominant name on the Helsinki exchange. Economists came to call the problem “the Nokia risk”: the danger that a small, wealthy economy stakes too much of its future on one champion. When Apple’s iPhone arrived in 2007 and Nokia’s handset business collapsed, the damage was national. From 2008 to 2019, real Finnish per-capita income declined, and a decade-plus of stagnation set in.

South Korea is now running the same play at national scale, with the bet on AI memory instead of mobile phones. The strength is real: SK Hynix is a leading supplier of the HBM that Nvidia and the hyperscalers depend on. But strength built on a single product is also fragility wearing a crown. Finland did not fail because Nokia made bad phones for most of its run. It failed because it had no cushion when the one thing it sold went out of fashion.

This Is Volatility, Not Yet Collapse

Now the honest counterargument. The cynical read is only half the truth.

The June crash was a valuation and positioning event, not a demand event. AI memory demand is still climbing, not falling. Broadcom’s own guidance points up, not down, and even Bloomberg, reporting through the selloff, noted that AI demand was beginning to justify the massive cost of the data center buildout. The market action backs this up: after the 10 percent plunge, the KOSPI rebounded over the following two sessions, the kind of whipsaw that signals forced selling and fear rather than a broken business. Korean stocks have been on a wild ride, not a one-way collapse.

So the synthesis is this. Korea’s bet on HBM is fundamentally sound. The product is real, the lead is real, the demand is real. What is dangerous is the concentration around it. When half your index is two stocks tied to one cyclical product, you do not need demand to collapse to get hurt. You only need demand to decelerate, or sentiment to wobble, and the whole market lurches. The bet is right. The portfolio construction is reckless.

What Comes Next

Watch the capital flows. Even as the home market wobbled, SK Hynix moved to raise about $29.4 billion through a new US listing. Read generously, that is a healthy company diversifying its investor base away from a single overheated exchange. Read skeptically, it is an offer to sell more of the concentration risk to American buyers near the top of the cycle. Both readings can be true at once.

The deeper issue is structural and will not resolve in a week. A market this concentrated invites fast, jittery money, and fast money leaves fast, putting pressure on the won and boxing in the Bank of Korea. The fix is not to wish away the AI boom that made Korea rich. It is breadth: more listed champions, deeper domestic demand, an index that is a basket again rather than a single wager. Until then, every Federal Reserve headline and every AI-spending jitter will keep hitting Seoul several times harder than it hits New York.

The Bottom Line

A 2 percent American sneeze gave Korea a 10 percent fever, the result of building a market on two chips and one bet. The technology is sound and the demand is real, which is exactly what makes the concentration so seductive and so dangerous. Finland already ran this experiment with Nokia and paid for a decade. The chips have changed. The risk has not.