Hong Hui Analyzes the AI Market Bubble: Why South Korean Semiconductor Stocks Collapsed So Rapidly

Why did the AI sector suddenly cool down in mid-year?

Why did the South Korean semiconductor market experience such a sharp turnaround from boom to bust?

Why should you be highly cautious when it appears that all investors are making money?

Recently, renowned economist Hong Hao shared his insights into the trends and underlying nature of AI market movements over the past year, along with his latest analysis. His discussion includes profound analyses of market dynamics and forward-looking perspectives worth noting.

Key quote:

1. The similarity between the current South Korean semiconductor stock rally and previous bubbles lies in the rapid rise of price bubbles due to uncontrolled investor leverage, which inevitably leads to a swift collapse of the bubble.

2. Behavioral economics tells us that most people become overwhelmed by large amounts of information and resort to using so-called "rules of thumb" to make decisions.

3. The efficient market pricing hypothesis, at first glance, is clearly invalid; moreover, the real world is likely the exact opposite of these assumptions.

4. Gold and other precious metals rising alongside stocks has only occurred before during the late 1970s and early 1980s—the period following the collapse of the Bretton Woods system.

5. When all investors use the same factor for asset allocation, it inevitably leads to self-reinforcing trends.

6. In a long-term market environment with suppressed volatility, investor leverage will inevitably increase.

7. As an investor, going against the market can be extremely exhausting. And before the bubble bursts, you’ll face all kinds of criticism and humiliation.

8. You might think buying thirty or forty stocks constitutes diversification, but if all of them are exposed to the same sector, you’re actually increasing your portfolio’s volatility.

9. Bubbles are the most powerful tool for transforming human society; imagine if there had been no AI bubble—it would have been difficult to mobilize so much capital to invest in our future.

Strictly applying historical patterns to a world that is beginning to change will lead to significant problems.

The most significant event in recent capital markets has been the surge, peak, and subsequent collapse of the global semiconductor sector. Led by South Korea, the semiconductor sector dropped roughly 50% from its peak within just 40 days.

The market is still experiencing significant volatility. Friends holding positions in South Korean semiconductor stocks are surely feeling immense stress, especially those using leverage, who face the constant threat of liquidation amid sharp price swings. I believe this is the most challenging period to endure.

This market rally resembles past bubbles in that price bubbles rose due to uncontrolled leverage by investors. However, any rise in the bubble line inevitably leads to a fall. Now we see that both the rise and fall of the bubble line have occurred without a prolonged process.

After reaching a peak near the 9,000 level, South Korea’s stock market sharply declined to around the 5,000 level before beginning to rebound. This market movement offers valuable insights for those who participated in the semiconductor rally. How to apply these lessons to our future predictions about markets, cycles, and bubbles is the key focus I want to emphasize today.

In traditional economic assumptions, there are many premises. We all know that in business school, we learn many models—such as the efficient market hypothesis, random walk, and expectation models—but these models rely on several critical assumptions.

First, assume that all market information is completely transparent and readily available to the general public. Clearly, this assumption does not hold. Anyone who has traded stocks knows that stock markets are filled with all kinds of negative news, and often, insider information is not reflected in stock prices. Yet, someone must have known first—after the news breaks, we frequently observe abnormal price movements before the official announcement, meaning the insider information had already been leaked in advance.

Second, assume that all market participants are perfectly rational. What does that mean? A rational person is highly sensitive to unit returns versus unit costs; when unit returns exceed unit costs, they automatically close their position and exit. At the same time, they are individuals seeking to maximize profit, equipped with a quantitative model that allows them to instantly process information and determine the optimal position. This is clearly even more implausible. Behavioral economics has shown that, in the face of massive amounts of information, most people’s decision-making processes collapse—their minds become overwhelmed, leading them to rely on so-called “rules of thumb,” which are often wrong when markets reach their extremes.

Third, assume there are no market frictions and zero transaction costs. We all know that transaction costs are substantial, especially as the size of the fund you manage increases—each trade exerts a greater impact on the market. In Hong Kong, a daily trading volume of just a few million yuan might be enough to control a stock; here on the mainland, I believe there are many micro-cap stocks. As a result, a popular strategy among many quantitative funds is to go long on a basket of small-cap and micro-cap stocks while shorting the index, aiming to capture what is called alpha. However, this alpha may very well be merely liquidity-driven.

Therefore, these so-called efficient market pricing hypotheses are clearly invalid—not one of them holds true. Moreover, the real world is likely the exact opposite of these assumptions.

In November last year, the entire market experienced a situation unseen in the past 60 years: gold and stocks rose together.

Gold has been surging at a parabolic pace, while stocks—whether U.S., emerging markets, or European—are also rising rapidly. This is unusual, because gold is traditionally a safe-haven asset; within a portfolio, its price movement is uncorrelated with other assets, automatically acting as a hedge when stocks plummet.

Now, not only gold but other precious metals are rising alongside stocks—a scenario last seen in the late 1970s and early 1980s, when the dollar-based credit currency system was taking shape and the Bretton Woods system collapsed. When gold and stocks rise together, it signals that the market has entered a regime shift, and if your portfolio lacks any hedging tools, it will be highly vulnerable when risks materialize.

Traditional portfolio management theory assumes that combining assets with no correlation can hedge against volatility. However, correlations inevitably change at market turning points or during shifts in market patterns—especially when risks concentrate and erupt. At such times, the correlations among all assets tend to move toward unity, causing the portfolio to lose its traditional hedging effect.

Over the past 40 years, China’s accession to the WTO and surge in labor productivity led to a systemic decline in global inflation. As a result, every market crash presented a buying opportunity—because declining inflation gave the Federal Reserve and other central banks room to continuously lower interest rates, which fell from double digits in the 1980s to single digits in the 2000s, reached zero during quantitative easing in 2008, and even turned negative in Europe. For the past four decades, we could safely assume that correlations among asset prices remained relatively stable. But all of this changed after 2020.

After 2020, quantitative easing intensified further—not only did central banks loosen monetary policy, but the U.S. Treasury also printed money and distributed it directly to American households, leading to an oversupply of currency. Look again: U.S. inflation reached as high as 9.1% in 2022, a level unseen in over 40 years.

The financial industry is extremely demanding; typically, being in the field for over a decade by your early 30s is considered long tenure, and analysts over 40 are extremely rare—those who make it to MD level are even fewer. Imagine when inflation surged above 8% in 2022, most analysts had never experienced such high inflation before. As a result, during this market regime shift, a wave of professionals entered who had never witnessed different market dynamics, and they continued using the same 60/40 equity-bond portfolio model they’d relied on for the past 14 years—only to be completely wiped out by 2022.

This is a major problem: rigidly applying the patterns of modern history to a world that is beginning to change.

There are several effective trading strategies in the market. One is momentum trading, which simply means the tendency for strong performers to continue performing strongly. Generally, the price momentum effect can accommodate only a limited amount of capital, but as far as I know, the vast majority of quantitative funds use trend-following strategies. As a result, when everyone allocates based on the same factor, it inevitably leads to self-reinforcing trends. Traders who employ this approach are called convergence traders and represent the market mainstream. Trend-following strategies remain effective until trends reach extreme levels.

Over the past two years, following the 2024 quantitative collapse, these quantitative strategies were highly effective until the first quarter of 2025 and 2026, when they largely failed by the second quarter.

Another type is the contrarian trader (also known as a divergence trader), who firmly believes in mean reversion—that cycles move from lows to highs and back to lows, typically taking three to four years. So, every two to three years, a quantitative value collapse occurs—simply put: after a time window opens and trends become extreme, mean reversion is inevitable.

Unfortunately, the market's design inherently encourages everyone to be a momentum trader, promoting following the trend rather than acting as a true critic. Going against the market is exhausting; moreover, before the bubble bursts, you’ll face endless criticism and abuse—because you’re standing in the way of others making money.

Another type of trader has no strategy—called a random trader—who has no model or consistent way of thinking; they buy when prices rise and sell when they fall. The mere presence of such random traders in the system inevitably leads to disorderly price expansion; their combined effect with convergence traders inevitably causes asset prices to become prone to bubbles. This is one of the reasons we’ve seen the current semiconductor bubble form.

Here’s a simple example: Some retail investors don’t know which sectors are good or bad—they simply buy what others are buying and sell what others are selling. The question is—have you ever considered why retail investors often end up buying at the top?

Many people say, “You called the South Korean semiconductor bubble correctly. So what?” They went from 4,000 to 9,000 and back down to 5,000, still making a 1,000-point profit. But don’t forget—very few were able to buy at 4,000; retail traders almost certainly bought at the top. Therefore, trading volume tends to concentrate at market peaks, while buying activity is lowest at market bottoms. Clearly, the rational trader assumed by the efficient market hypothesis doesn’t exist.

We often talk about a “slow bull market,” hoping the market rises gradually. But consider this: if the market truly experienced a slow bull, growing at 8% annually with minimal pullbacks, while the 10-year interest rate was only around 1.8%, I would naturally be willing to borrow at 1.8% to invest in a market growing at 8%. Such a scenario would inevitably lead to increased market leverage—or, in other words, leverage will inevitably rise in a market with suppressed volatility.

Where does the leverage come from? Currently, hedge funds on both the long and short sides are increasing their leverage. In Korea, leverage stems from the issuance of leveraged ETF products, such as the 7709, a 2x long ETF on Hynix, which rose 12-fold in a year—but also caused massive losses and the highest number of liquidations among investors. Additionally, retail investors are using margin trading, financing, and derivatives.

In a market characterized by calm conditions but rising expectations of price increases, leverage is inevitably inflated. Meanwhile, South Korea approved the listing of leveraged ETFs in June, citing the fact that Hong Kong’s 2x long ETF had once been the world’s largest single-stock leveraged ETF; South Koreans felt, “We can’t let the Chinese make all the money,” and thus launched their own 2x and 3x long ETFs on SK Hynix and Samsung. Coincidentally, in June—when the market had already reached extremely high levels—leveraged instruments were unleashed at historical peaks, which is precisely a major trading mistake.

When a sector keeps doubling—Samsung and SK Hynix together make up about 20% of the Korean market—if they triple in a year, their weight will exceed 50%—if they continue rising, the entire Korean market would eventually consist only of Samsung and SK Hynix, which is absolutely impossible. Therefore, generally speaking, when an index rises two to threefold, it inevitably hits an upward ceiling. The current movement of Korea’s KOSPI mirrors the historical pattern of tech stock bubbles; it cannot possibly triple in a single year, or there would be no room left for other stocks to survive. When Samsung and SK Hynix accounted for over 50%, nearing 60%, of the Korean market, they collapsed dramatically.

The calmness of the entire market, combined with the use of leverage, directly fueled the creation of leverage; as leverage increased further, prices rose at an accelerating pace, and this price appreciation itself provided a reason for trend traders and momentum traders to buy—rising prices were justification enough to buy, simply that, until the bubble burst. The same happened in 2026.

Three necessary conditions are required for a bubble to form.

First, the grand narrative: Technological progress generates grand narratives; the more advanced and revolutionary the technology, the stronger the narrative becomes, increasing the likelihood of a bubble—people’s expectations are stirred, and everyone believes this wave of technological revolution will lead to increased labor productivity and income growth. Society becomes captivated by the grand narrative, fostering boundless optimism about the future.

Second, the use of leverage, or the relaxation of regulations.

Third, investors' profit-seeking behavior—driven by high expectations for the future and easy access to funding—leads everyone to chase price increases. Thus, the more grandiose the narrative, the more capital available, and the calmer the market, the more likely a bubble is to form.

Those involved in insurance pricing should know that before the conflict, insurance premiums for ships passing through the Strait of Hormuz remained stable—tens of thousands of dollars per vessel—because everyone assumed the route had been safely navigated since the 1990s, so they used historical pricing models to set current rates. But anyone in insurance pricing knows: the calmer the conditions and the lower the premiums, the more vigilant you should be, because risk can emerge at any moment.

The market today is the same: when traders at all levels align their expectations, bubbles and leverage inevitably emerge; those who dare to go against the cycle, voice dissent, and remain清醒 amid the torrent—I believe by June, there will hardly be any left—this is precisely the most dangerous state of a bubble.

When we diversify our investments, we often think that buying a basket of stocks is enough to achieve diversification. Sometimes I see friends holding thirty or forty, or even more, different stocks—and when I ask what each one does, most can’t answer: they don’t know what Company X does, or what Minimax does; they only know it’s a leading AI company. But the risk exposure for all these thirty or forty stocks is essentially the same—they’re all exposed to fluctuations in the semiconductor sector, creating a shared risk. As a result, even with thirty or forty stocks in your portfolio, you don’t reduce—and may even increase—your portfolio’s volatility, because your chances of losing money have grown.

It’s like a forest: a sparse, young forest has very little risk of wildfire; but once it becomes dense and mature, with all the large branches intertwined, a single lightning strike can destroy the entire forest. So don’t assume that during times of peace, when trees grow vigorously, peace has been established and risk has disappeared—in fact, the more peaceful the period, the higher the risk.

This round of the Korean bubble collapsed faster than any recorded in human history—half of it vanished in just 40 days. If you held a 7709 2x long position in Hynix, the price dropped from 200 to 20—a 90% decline—then surged 70% in a single day. The bubble has now entered a consolidation and digestion phase, with the typical correction range during a bubble collapse falling between 50% and 65%, or roughly one-half to two-thirds.

Larger stock markets, such as Nasdaq, experienced a peak decline of approximately 80% in March 2000. If you were a buy-and-hold long-term investor, it took more than 20 years before the market regained its March 2000 high.

So some people say, no matter what, buying XX shares will make you money; but what matters is the sequence in which you do it.

But believe me, this will still happen, because bubbles are the most powerful tools for transforming human society. Without such a massive bubble, it would be difficult to mobilize so much human and material capital to bet on humanity’s future—investing in AI devices and building cloud computing centers.

The future world has already taken shape: we need more GPUs for inference computing, more storage, and more materials to build robots. It will not stop because of a burst AI stock bubble; instead, new models, new investments, and new opportunities will emerge from the ruins, shaping an even brighter future.

This article is from the WeChat official account "Capital Deep Dive," authored by Capital Deep Dive.