Korea Is the Cheapest Stock Market in Twenty Years. That's Exactly the Problem.

A major institutional asset manager just called Korea "off the chart cheap." Korea's stock market is also up 156 percent over the past twelve months. Both facts are true. The trick is in how you measure cheapness.

A major institutional asset manager published a research note last week with a striking centrepiece chart. On that chart was a measure of valuation for the world's major equity markets, plotted against twenty years of history. Most countries sat near the middle of their historical range. The United States, expensive. Japan, expensive. Even Taiwan, after a screaming year, was trading near the top of its range.

One country sat at the very bottom. Not just the bottom of the chart, but the bottom of its own twenty-year history. The cheapest it had been in two decades.

That country was South Korea. The report's authors called it "off the chart cheap."

And here is the problem with that statement: Korea's stock market is also up 156 percent over the past twelve months. The KOSPI closed last week at 6,475, having started the year around 4,200. Samsung Electronics and SK Hynix — the two companies that, between them, account for nearly a third of Korea's entire market capitalisation — just delivered the highest combined quarterly profit in the country's corporate history.

KOSPI Composite Index · Daily · Korea's benchmark has rallied from ~4,200 in December 2025 to 6,651 — a 156% gain in twelve months. The February consolidation broke higher in April. This is the stock market that "off-the-chart cheap" is supposed to describe.

Both things are true. The market has rallied harder than any in the world. And by the most popular yardstick used by professional investors, it is somehow simultaneously the cheapest market on Earth.

When two facts like that sit next to each other and don't seem to make sense, one of them is usually misleading. In this case, neither is. The trick is in how you measure cheapness.

The Metric That Lies When You Need It Most

Let me explain the yardstick, because it has a name and a specific weakness, and once you see how it works you cannot unsee it.

It is called PEG, which stands for price-to-earnings-to-growth. The arithmetic is simple. You take a company's price-to-earnings ratio — how much you are paying for each dollar of profit — and divide it by how fast its earnings are growing. If a stock trades at twenty times earnings and its profits are growing at twenty percent a year, its PEG is one. Below one, the metric calls a bargain. The lower the number, the cheaper the stock is supposed to be.

PEG was popularised in the 1980s by Peter Lynch, the legendary fund manager who ran Fidelity's Magellan fund. It became one of the most widely cited valuation tools in the world, taught in business schools and quoted in research notes everywhere.

What most people who use it have forgotten — or never knew — is that Lynch himself warned, repeatedly, that PEG does not work on cyclical companies. He said this directly in his books. The metric was designed for steady, predictable growers — companies that compound at a similar rate year after year. For those, PEG is a useful shortcut. For commodity producers, automakers, airlines, and especially semiconductor makers, Lynch warned, PEG would tell you exactly the wrong thing at exactly the wrong time.

Here is why. Look at the math: price divided by earnings, divided by growth. If growth is huge, the answer comes out small. The metric flashes "cheap." But growth is highest at the top of a cycle, when business is roaring, prices are rising, and demand is at its peak. That is also the moment the cycle is most likely to turn. Order books look full because everyone has been ordering ahead of feared shortages. Margins look fat because supply has not yet caught up to demand. It looks like a permanent boom. It always does.

Then supply catches up. Inventories build. Customers stop ordering because they already have too much. Prices drop. Margins collapse. Earnings fall fifty, seventy, ninety percent, sometimes turning into outright losses. And the PEG ratio, which had been screaming "buy," resets — but by then the stock is already down.

This is not a theoretical concern. It is what has happened, on a near-decadal rhythm, in the very industry that drives Korea's stock market.

Three Booms, Two Busts

The Korean stock market is not a balanced index. It is, to a remarkable degree, a bet on two companies — Samsung Electronics and SK Hynix — and within those companies, a bet on one product: memory chips. The DRAM and NAND that go into your phone, your laptop, the data centres that train artificial intelligence models. When memory prices rise, Korean profits explode. When memory prices fall, they implode.

Memory chips are perhaps the most cyclical commodity on Earth. They follow a predictable, brutal pattern: demand outpaces supply, prices rise, the chip companies make fortunes, they spend those fortunes building new factories, the new capacity comes online two years later, supply now exceeds demand, prices crash, profits vanish, factories sit idle. Then demand catches up again, and the cycle repeats. It has worked this way for forty years.

In 2017 and 2018, the world experienced what was called the "memory super-cycle." Samsung's semiconductor division and SK Hynix together earned roughly 80 trillion Korean won in operating profit in 2018 — at the time, the largest combined number these two companies had ever recorded. Research notes flooded out calling Korean equities "exceptionally cheap on PEG." The denominator — earnings growth — was so large that the metric showed a bargain even though stock prices had doubled.

In 2019, the cycle turned. Combined operating profit fell roughly 79 percent. From 80 trillion won to 17 trillion. The KOSPI dropped about 23 percent over the following twelve months. Investors who had bought the "cheap on PEG" thesis were the bag holders.

The market recovered through 2020 and 2021, this time pushed by remote-work demand and the early stages of the AI boom. Combined chip profits hit 58 trillion won in 2021. Once again, the research notes appeared. Once again, PEG looked like a screaming buy. Once again, investors piled in.

In 2023, the cycle turned harder than almost anyone expected. Samsung's chip division lost roughly 15 trillion won. SK Hynix lost 7.7 trillion won — the largest annual loss in its history. Combined, the two companies' chip operations went from making 58 trillion won to losing 22 trillion won. A 137 percent swing in three years. The KOSPI fell about 29 percent over fourteen months from its peak.

You might be starting to notice a pattern.

The Biggest Peak Yet

We are now in the third peak. And it dwarfs the previous two.

In the first quarter of 2026 alone — three months — Samsung's semiconductor division delivered 53 trillion won in operating profit. SK Hynix delivered 38 trillion. Combined: 91 trillion won. In a single quarter. SK Hynix's operating margin hit 72 percent, surpassing TSMC for the first time in seven years. The full-year combined chip profit is on track to exceed 350 trillion won, more than four times the 2018 super-cycle peak that the chart still shows as a tower.

The driver this time is artificial intelligence. Specifically, a type of memory called HBM — high-bandwidth memory — which sits next to the chips Nvidia sells to OpenAI and Google and Meta. Every AI data centre needs enormous amounts of HBM. Korea makes most of it. Demand has wildly outstripped supply. Some commodity memory contract prices have risen 100 percent in a single quarter. Samsung and SK Hynix have asked customers to pay 70 percent more for server memory in the first quarter of 2026, and customers paid it.

This is exactly when PEG looks cheapest. Earnings growth is being measured against last year, when the industry was still recovering from the 2023 crash. The denominator of the ratio is enormous. The metric flashes "off the chart cheap."

But ask the question Peter Lynch would ask. At what point in the cycle does memory pricing rise 100 percent in a single quarter? At what point in the cycle does an entire industry book record-shattering profits four times larger than the previous super-cycle?

This Time Is Different. (It Never Is.)

There is always an argument for why this peak is different. In 2018, it was that smartphones and cloud computing had created permanent new demand. In 2021, it was that the post-pandemic shift to remote work had structurally changed memory consumption. Both arguments turned out to be wrong, or at least premature.

The argument now is that AI demand is not a cycle but a structural shift — that data centres being built today represent a decade-long build-out and memory undersupply will persist for years. Nomura forecasts the industry remaining tight through 2028. Some analysts argue HBM is so technically demanding to manufacture that supply cannot easily ramp.

These arguments may turn out to be partly right. AI demand is real. The industry probably is in better structural shape than it was in 2018 or 2021. But "this cycle is bigger and longer than the previous ones" is a different claim from "this cycle does not exist." Cycles always exist. What changes is their magnitude and timing. The bigger and faster the boom, the bigger the inventory build, and the bigger the eventual crash.

There is also a tell that should worry anyone watching closely. Korean retail investors — who have ridden Samsung and SK Hynix all the way up — sold a net 5.6 trillion won of these two stocks between April 7th and 24th. Foreign institutional investors bought a net 1.7 trillion won over the same period. Locals are taking profits. Foreigners are still chasing.

What This Means If You Are Reading From a Distance

If you are a retail investor in Malaysia, or anywhere outside the institutional world, here is what you need to take from this.

When you see headlines or research notes calling Korea "the cheapest stock market in twenty years," understand what is actually being measured. It is a metric called PEG, and it is being applied to an industry — semiconductor memory — for which its inventor explicitly warned it does not work. The metric is reflecting the math correctly. The math is the wrong math for this situation.

Understand that the "cheapness" is being driven entirely by an extraordinary, possibly unrepeatable burst of earnings. If those earnings simply normalise — not collapse, just slow down — the apparent cheapness disappears even if the stock price does not move. If the earnings actually crash, the metric will flash "expensive" right after the stock has already fallen thirty percent.

Understand the historical record:

In the past decade, Korea's stock market has had three earnings peaks and two earnings crashes. Each peak produced the same valuation argument. Each crash produced 20 to 30 percent drawdowns in the index. The current peak is more than four times the size of the previous record. The historical pattern says that the eventual reversion is likely to be larger, not smaller.

This is not a prediction that Korea crashes next month, or next quarter. The market might keep rallying. AI demand is real. The cycle could extend further than past patterns suggest. Sometimes the seventh inning lasts a long time.

But buying because something is "cheap on PEG" — when you are at the largest cyclical earnings peak in the history of one of the most cyclical industries on Earth — is not a value purchase. It is a momentum bet wearing the costume of value investing. And the most expensive mistakes in markets, throughout history, have been made by people who confused the two.

The asset manager that called Korea "off the chart cheap" is paid to allocate institutional money. They will be fine. Their fund will be measured against a benchmark, and if Korea falls, the benchmark falls with it, and they will look like everyone else.

You will not be measured against a benchmark. You will be measured against your own savings.

Choose accordingly.