NextFin News - South Korean stocks staged a violent rebound on Friday, with the Kospi climbing as much as 14% and SK Hynix jumping as much as 28% after a week of sharp losses, as SK Hynix’s record second-quarter earnings and a rare open-market share purchase by SK Group Chairman Chey Tae-won revived confidence in the AI memory trade. Samsung Electronics rose as much as 26% in the same session, underscoring how tightly the market has become tied to the fortunes of a handful of AI-linked chipmakers.
The Rebound Was Less a New Narrative Than a Forced Repricing
As of the July 30 Seoul session and the July 31 UTC publication window, the market’s message was clear: the selloff had gone far enough to make confidence itself tradable. The first question is not why Korean stocks bounced, but why the rebound was so abrupt. The answer begins with a market that had already been punished hard, then received two forms of reinforcement within 24 hours: SK hynix’s record second-quarter earnings and Chey Tae-won’s first direct open-market purchase of the stock.
SK hynix said in its official 2Q26 release that it recorded revenue of 79.3187 trillion won, operating profit of 60.5426 trillion won and net income of 93.9226 trillion won, all record highs. Revenue rose 51% from the prior quarter and 257% from a year earlier, while operating profit rose 61% sequentially and 557% year on year. The company also said its first-half revenue crossed 100 trillion won for the first time and that HBM4 mass shipments began in the second quarter. Those numbers do not describe a business losing traction. They describe a business still being pulled upward by AI-related demand faster than supply can normalize.
Chey’s purchase then turned that earnings strength into a market signal. SK hynix said in a regulatory filing that he bought 3,620 common shares in the open market, worth about 4.8 billion won at the stock’s closing price of 1.32 million won on July 30. Yonhap reported that the purchase was his first direct buy of the stock. The quantity was small relative to the company’s market value, but the message was large: the chairman chose to buy while the stock was under pressure rather than simply say the business was undervalued.
That combination matters because it changes how traders interpret the same earnings data. Alone, the quarter would have been read as strong. Alone, the purchase would have been a confidence gesture. Together, they forced a reassessment of whether the recent selloff had become detached from the underlying earnings engine. When a high-conviction market is already fragile, a record print and a visible insider-style vote of confidence can trigger more than a bounce; they can trigger a squeeze in the risk premium itself.
The key point is that Friday’s move was not just a one-day relief rally. It was a repricing of confidence in the duration of the AI memory cycle after a week in which traders had begun to question whether the trade had become too crowded and too fragile. The market had sold first and asked questions later. Chey’s purchase and SK hynix’s earnings forced it to ask those questions again.
Why The AI Trade Still Runs Through A Few Korean Names
The deeper mechanism is structural, even if the day-to-day price action is cyclical. Korea’s market is not rising because investors suddenly discovered semiconductors; it is rising because the country has become a concentrated equity expression of global AI capex. When hyperscalers expand data-center spending, the transmission channel runs through accelerators, then into high-bandwidth memory, then directly into SK Hynix and Samsung Electronics. That makes Korean equities less a local story than a leveraged derivative of global AI infrastructure spending.
That transmission is visible in the company’s own language. SK hynix said in its second-quarter release that the quarter reflected “strong demand growth from expanding AI infrastructure investments” and noted that it had begun mass shipments of HBM4 in the period. The company also said cumulative first-half revenue crossed 100 trillion won for the first time. Those details matter because they show demand is not confined to one product cycle; it is broadening across premium DRAM, HBM and enterprise storage.
“strong demand growth from expanding AI infrastructure investments”
That phrase is the key to the whole trade. It suggests the market is not just betting on a temporary inventory restock. It is betting that AI compute build-outs keep pulling on memory demand even as prices and margins move higher. In other words, the stock is not only reacting to quarterly earnings. It is pricing a multi-quarter capacity and pricing cycle where the supply side remains tight and the product mix keeps moving up the value chain.
But this is where the second-order question matters. If the market has already decided AI capex is durable, why did the stock still swing so violently? Because the margin for error in a concentrated market is thin. When Samsung Electronics and SK Hynix together account for a very large share of the Kospi’s weight, every swing in their prices feeds back into the index, into retail sentiment, and then into foreign flows. That creates a loop: higher AI optimism lifts the chips, the chips lift the index, the index validates the optimism, and then the market becomes more sensitive to any sign that the loop might break.
This is why the Korean rebound is best understood as both cyclical and structural, but at different horizons. The short-term move is cyclical. It is a sentiment bounce after a violent drawdown, amplified by a chairman purchase and a strong earnings print. The longer-term setup is structural. Korea has become one of the clearest public market vehicles for AI infrastructure spending, and that role will not disappear simply because the share prices are volatile.
The distinction matters. Cyclical rebounds mean-trade and fade. Structural repricings persist until the underlying industry architecture changes. Here, the architecture is still intact: cloud capex is large, HBM demand remains tight, and the Korean memory oligopoly still sits in a privileged place in the supply chain. That does not make the shares risk-free. It makes the volatility a feature of the regime, not an argument against the regime itself.
It also helps explain why the broader market can rise so much when the same two names move. In a normal index, a single company’s earnings beat changes one stock’s valuation. In Korea’s market, the same beat can change the index’s perceived direction, and the index can then change the signal that global investors read about the AI cycle. That feedback loop is why Seoul increasingly trades like an early read-through on semiconductor risk elsewhere.
There is a second structural layer as well. SK hynix’s release noted that it had begun mass shipments of HBM4 and was already selling into long-term customer demand. That matters because product transitions usually reset market leadership. Here, the transition appears to be reinforcing leadership instead: the firm is moving into the next generation of AI memory while demand is still strong for the current one. That is a better setup than a simple late-cycle price pop, because it suggests the company is not only riding the old wave; it is entering the next one with bargaining power intact.
The Bull Case Is Strong, But The Bear Case Starts At The Same Place
The strongest counter-thesis is that the rally is exactly what speculative bubbles look like: a management gesture, a hot earnings print and an AI slogan all arriving after a sharp selloff. On that reading, the market is not confirming durable strength; it is simply forcing a violent squeeze in a crowded trade. The memory cycle has always been cyclical, and every time investors have convinced themselves that a new technology abolished the cycle, the industry eventually reminded them that supply still comes back, pricing still normalizes and margins still compress.
That argument deserves respect because it attacks the thesis at its foundation. SK hynix’s outperformance can be read as a function of extreme concentration, not durable breadth. If AI spending slows, if HBM pricing cools, or if competitors expand supply faster than expected, the market can reverse quickly. The historical pattern for semiconductors is brutal: demand upswings look linear until they do not, and then prices and margins adjust far faster than investors’ models.
Yet the bear case is not enough to explain the day’s move by itself. The quarterly numbers were too large, too recent and too directly tied to AI infrastructure demand to be dismissed as a mere sentiment echo. Revenue of 79.3187 trillion won and operating profit of 60.5426 trillion won are not soft evidence. They show the company is extracting more value from the current build-out than a cyclical peak narrative would imply. The market can still overshoot, but it is harder to call a business structurally fragile when it is posting record revenue, record operating profit and record net income in the same quarter.
The falsifying signal is equally clear. If SK hynix’s next quarter fails to show continued HBM and high-end DRAM shipment growth, or if management warns that AI-related demand is slowing while margins compress materially from the 76% operating margin reported in the second quarter, the structural-bull view weakens sharply. For the broader Korea trade, a sustained reversal in the Kospi’s chip leaders would tell the same story even faster: if the leaders stop carrying the index, the narrative stops carrying the market.
For now, the burden of proof sits with the bears. They need to show not just that the stock has become expensive, but that the earnings and demand engine behind it has started to crack. Until that happens, the market is treating the recent volatility as a shakeout inside a still-powerful AI memory cycle, not the end of it.
The bearish case does have one more weapon: memory demand can stay strong while stocks still fall if valuations outrun the cash flow path. That is why the market’s next move will matter so much. If the share price keeps rising faster than the earnings base, the same structural story can still become a speculative one. The difference is not whether AI demand exists. It is whether the market is still willing to pay for it after another round of numbers.
What Comes Next Depends On Which Horizon You Watch
In the short term, the beneficiaries are clear. SK Hynix, Samsung Electronics and the broader Kospi benefit from renewed confidence in AI-related capex, while short sellers and late sellers are exposed to further squeezes if global chip stocks keep holding up. A continuation of the rebound would likely depend on follow-through buying in US AI peers and another round of confirmation that memory demand is still outrunning supply. If the US peers weaken again, Korea can give back gains just as quickly.
Medium term, the market is likely to keep using SK Hynix as a proxy for AI infrastructure spending, which means earnings reports and customer commentary will matter more than headline index moves. The next clues to watch are shipment commentary on HBM4, pricing trends for advanced DRAM and the company’s tone on capacity expansion. If those indicators remain firm, the rebound has room to last as a valuation and earnings story. If they soften, the rally looks more like a tactical unwind than a durable trend.
Long term, the structural takeaway is that Korea’s market has become unusually dependent on a narrow set of AI-linked names. That concentration can keep generating outsized gains when the cycle is favorable, but it also makes the index vulnerable to any break in AI capex or memory pricing. The very mechanism that lifted the market Friday can cut the other way when the industry turns.
The base case is a volatile but still supportive AI memory cycle, with periodic air pockets and sharp rebounds as investors recalibrate. The upside case is that HBM demand stays tight longer than expected and keeps re-rating both SK hynix and Samsung as earnings visibility improves. The downside case is a faster-than-expected supply response or a slowdown in hyperscaler spending, which would expose how much of Korea’s rally rests on a single trade.
Over the next one to two quarters, the test will not be whether AI remains important. It will be whether SK hynix can keep translating that importance into shipment growth and margin durability while the market keeps paying up for the same story. If it can, Friday’s rebound will look like a reset inside an advancing cycle. If it cannot, the day’s move will be remembered as the moment traders confused a relief rally with a regime change.
Friday’s bounce showed that the AI story in Korea is still alive. It also showed how little room there is for doubt once the trade becomes the market.
Explore more exclusive insights at nextfin.ai.
