On the other side of the world, in a country the size of the state of Indiana (physically, not population), something interesting is happening. But before we transport ourselves to South Korea, we need to take a quick detour. After Wednesday’s close, Micron (a Boise, Idaho memory chip company) reported an absolutely massive quarter. They did $41.5 billion in revenue (more in three months than in all of 2024), with a gross margin north of 80%. That is the kind of figure you associate with software, not with memory.
But while Micron is the name we hear about most, being U.S.-based, it is actually the smallest (by market share) of the three main players that make DRAM (dynamic random access memory), and more importantly, HBM (high-bandwidth memory), which is a stacked DRAM that sits alongside almost every AI accelerator Nvidia ships. Which raises the question of what that print meant for the two players running the show in, you guessed it, South Korea.
However, what you may NOT have guessed is that they also happen to be OVER half of the entire national stock market. And when Seoul opened Thursday, the KOSPI did not brush off the news. It leapt 5.4% in a single session to close just under 9,000, because the index had just received the not-so-subtle message from Idaho: the memory shortage is real, the pricing is holding, and your two crown jewels should be performing the same way. Well, at least that is the narrative.
From Sleepy to Supersonic
For most of its existence, the KOSPI has been, shall we say, unremarkable. First reaching 1,000 in the mid-90s, it spent decades as an unloved emerging-market benchmark. It first reached 2,000 in 2007; however, it was only at 2,200 in early 2020 before the onset of COVID. And while it recovered post-COVID, by the end of 2024, it could only maintain 2,400. I won’t make you do the math, but it returned about 3.5% per year for 35 years. Ouch.
But then, something happened. In 2025, the index returned roughly 75%, making it the best-performing major market in the world. But that turned out to be just a warm-up. Through the first half of 2026, the KOSPI cleared 5,000 in January, 6,000 in February, blew past 7,000 and 8,000 in the spring, and touched an intraday peak above 9,385 in mid-June. Year-to-date, the index has more than doubled. The pace is now openly compared to the Nasdaq’s 102% run in 1999, a comparison that is either thrilling or terrifying, depending mostly on your personal risk tolerance.
But the ride has not been smooth. In the first half of 2026, the KOSPI closed up or down 5% or more on twenty separate trading days. In all of 2025, it did that twice. This is an index of trillion-dollar companies now moving like a basket of meme stocks, and the reason why is the most interesting part.
Two Stocks in an Index’s Clothes
Here is the part that should make any student of diversification sit up straighter. The KOSPI tracks over 800 companies. But as mentioned earlier, Samsung Electronics and SK Hynix together now account for over half of its market capitalization. And according to one bank’s accounting, those two names have driven about 70% of the KOSPI’s gains this year. This is no longer a broad stock market. It is a bet on two memory companies wearing an index’s clothes.
And the reason (for now) is genuinely fundamental, not just froth. High-bandwidth memory is currently the bottleneck in the AI build-out. Every Nvidia chip needs a stack of it sitting inches away, and there is nowhere near enough to go around. SK Hynix, a company nearly bankrupt with $14 billion of debt a little over a decade ago, is now the undisputed king of HBM, with close to 60% of the global market and the lion’s share of Nvidia’s next-generation orders locked up. On June 22, in a moment that reorders 27 years of Korean corporate history, it overtook Samsung as the country’s most valuable company for the first time. Samsung had held that crown for 25 years and 7 months. It changed heads because of stacked memory dies.
The financials underneath are no joke. SK Hynix posted a quarterly operating margin of around 72%, more profitable as a percentage than Nvidia itself. When the component supplier out-margins the company selling the finished superchip, you are watching a textbook supply squeeze in real time.
The 15,000 Question
This is the backdrop against which JP Morgan just did something remarkable. The bank lifted its 12-month base-case KOSPI target to 12,500, kept Korea as its single most preferred market in Asia, and set a bull-case target of 15,000. For an index that closed 2025 around 4,214, a 15,000 print would be a nearly 3.5x increase in 18 months.
The thinking is based on operating leverage. Memory manufacturers carry enormous fixed costs: the fabs are built, the equipment bought, the engineers paid. Once that base is covered, additional revenue from higher chip prices falls almost straight to the bottom line. Goldman Sachs, which carries its own 12,000 target, forecasts 2026 earnings growth for the KOSPI of around 300%, which it notes would be the strongest annual profit expansion for any Asian market since the late 90s. The bull case is not really a bet on the stock market going up, though; it is a bet that AI memory demand will keep outstripping supply, and that those fat margins will keep compounding into 2027.
The ‘Ants’ Go Marching
Now, some people see this as a cause for concern, particularly for regulators, who see the structure stop being a feature and becoming a bug.
Korean retail investors, affectionately known as “ants,” have not merely participated in this rally. They have marched millions strong into it. The country now has more than 105 million active brokerage accounts, against a total population of 51 million, which is roughly two accounts for every man, woman, and child. These account holders have redirected savings into equities and margin debt, which recently breached a record 38 trillion won, up from 27 trillion at the end of 2025. Roughly 9 trillion won of it is concentrated in just two names. You know which two.
Then there are the leveraged ETFs. In late April, regulators approved single-stock 2x leveraged funds tracking Samsung and SK Hynix. The result has been spectacular and slightly alarming. These products now account for something like 31% of Samsung’s daily trading volume and 38% of SK Hynix’s. For every ten shares of SK Hynix that change hands, nearly four are tied to leveraged vehicles that did not exist a few months ago. The head of Korea’s Financial Supervisory Service publicly said he regrets not blocking them.
Why the regret? Because leverage is symmetric, and gravity is patient. Leveraged ETFs must rebalance daily, mechanically buying more as prices rise and selling as they fall. And on Tuesday, June 23, after the regulator’s comments rattled an already-jittery market, the selling cascaded. Samsung fell 12.3%, and SK Hynix fell 12.5% in a single session, dragging the KOSPI down 9.99% to 8,204 and tripping a market-wide circuit breaker. The leveraged ETFs tracking those two names fell roughly twice as far, around 24% to 26%. The market has since clawed most of it back within days, which tells you something about both the conviction and the fragility on display.
The Memory Chip Lesson
There is a temptation, watching a chart go vertical, to argue only about whether the story is real. The crazy thing about the KOSPI is that the story is real. The AI memory supercycle is not a meme. The margins are not imaginary. The 300% earnings growth is actually happening.
But a great story and a great investment are not the same thing, and conflating them is one of the most expensive mistakes an investor can make. Concentration and leverage are amplifiers, and they work in both directions. The very mechanisms that turbocharged the climb (two-stock concentration, record margin debt, and leveraged ETFs) are pre-wired to run in reverse the moment sentiment turns.
Also notice that even the bulls are quietly telling on themselves. JP Morgan’s bull case is 15,000. Its bear case, published in the same breath, is 8,000 (below where the index trades today). When your analysts hand you a range that wide, have they even made a prediction?
None of this is a call to short Korea or swear off the AI trade. It is a reminder that the right question is rarely the simple one. It is “what happens if I’m right, what happens if I’m wrong, and have I sized this so I survive the second one?” And in this specific case, “how long does the memory supply shortage last?” Because markets are forward-looking, what is happening today (and expected next year) is already priced into the stock. But at some point, when (not if) something changes, you’ll wish you had asked yourself, “What happens if I’m wrong?”
Markets / Economy
- It was a down week as market volatility picked up with the AI-chip trade swinging massively in both directions. The S&P finished the week down -2.0%, the Nasdaq down -4.6%, and the small-cap Russell 2000 up 1.0%.
- New home sales sank by 7.3% from the prior month to an annualized rate of 580K in May. It was the sharpest decline since January.
- Core PCE price index rose by 0.3% MoM in May, the same as an upwardly revised 0.3% in April and in line with market forecasts.
- Personal income increased by 0.7% MoM in May, surpassing market expectations of a 0.4% rise and accelerating from a flat reading in April. It marked the strongest monthly gain since July 2025.
Stocks
- U.S. equities were in negative territory. Technology and Communication Services led the decline, while Healthcare and Utilities outperformed. Value stocks led growth stocks, and small caps beat large caps.
- International equities closed lower for the week. Developed markets fared better than emerging markets.
Bonds
- The 10-year Treasury bond yield decreased eight basis points to 4.38% during the week.
- Global bond markets were in positive territory this week.
- Government bonds led for the week, followed by corporate bonds and high-yield bonds.
