Why Seoul's AI Stock rout matters beyond Korea

The honeymoon is over

South Korea’s market was the darling of the first half of 2026. The Kospi became the world’s best-performing major equity benchmark, powered by a single, soaring thesis: artificial intelligence demands memory chips, and South Korea makes the memory chips. Samsung Electronics and SK Hynix together command more than half the index. When AI capex looked infinite, that concentration was a feature. Now it is a liability.

The reversal has been swift and unambiguous. Trading turnover across South Korea’s $4.3 trillion equity market has collapsed 70% from its late-May peak. The Kospi has shed 22% since mid-year. Foreign investors have withdrawn $131 billion from Korean equities this year — the largest outflow among major Asian markets — according to Bloomberg-exchange data. That figure alone should give any portfolio manager sitting on AI-heavy EM positions pause.

What makes this selloff particularly notable is its breadth. This is not a single-name correction or a sector-specific pullback. The outflows have swept through the entire market, dragging down even companies with no direct AI exposure. When foreign investors flee a market, they tend to sell indiscriminately — liquidity comes first, fundamentals second. That dynamic is playing out in real time in Seoul.

A rally built on one pair of companies

The structural weakness here is not subtle. Two companies — Samsung and SK Hynix — dominate the index and the narrative. When those names falter, the entire benchmark follows. The market is now grappling with a contradiction that should have been obvious months ago: Samsung reported a nearly ninefold increase in quarterly operating profit, yet its shares still declined on the news. Profit growth alone no longer moves Korean tech stocks. Investors are pricing in the possibility that the AI memory-chip cycle may not extend as long as anyone hoped.

That pricing discrepancy reveals something important about the market’s psychology. When a company delivers one of the most dramatic profit surges in its history and the stock still falls, you are no longer trading on earnings — you are trading on expectations of what comes next. And right now, that forward look is uncomfortably narrow. The AI memory-chip supercycle narrative has dominated Korean equity strategy for eighteen months. If that narrative fractures, there is very little left to prop up a market that has priced itself as if HBM demand would be perpetual.

Corporate support is simultaneously evaporating. Samsung and SK Hynix are nearing completion of combined share-repurchase programmes worth 55 trillion won ($41 billion). Once those buybacks conclude — a floor that has underpinned sentiment — there will be little left to catch a falling knife. Those programmes were announced with the implicit promise that management believed the market had undervalued the company. The fact that they are being wound down while prices remain depressed sends a quiet but unmistakable signal.

Retail participation, which helped fuel the early-rally momentum, has also receded following July’s selloff. Outstanding margin loans have dropped to roughly 33 trillion won from a June peak of 38.6 trillion won. Brokerage deposits have fallen from nearly 140 trillion won to around 100 trillion won. The domestic liquidity engine that amplified the rally is now running dry. That is significant because Korean retail investors have historically been among the most aggressive buyers in emerging markets during bull phases. Their retreat is not a minor footnote — it is a structural shift in demand that removes one of the rally’s key supports.

Where the money is going instead

Capital is not disappearing from the region. It is rotating. Taiwan’s Taiex has gained 70% this year, outperforming the Kospi by roughly 23 percentage points last quarter alone. The rotation is not random. Taiwan offers a broader, more diversified exposure to the AI supply chain — from fabricators to design houses to packaging — rather than the concentrated memory-chip bet that defines South Korea.

UBS Global Wealth Management has explicitly shifted its tactical AI exposure toward Taiwan equities, citing the island’s diversified technology-hardware ecosystem and the sustained capital spending of major AI buyers. The logic is straightforward: if you want AI upside in Asia and you do not want your entire position dependent on HBM demand forecasts for Samsung and SK Hynix, Taiwan is the play.

This is the first-order implication that most English-language coverage is still missing. The Korean slowdown is not just a local story. It is a signal that even the most committed institutional buyers are beginning to scrutinize the sustainability of AI-driven earnings projections — and they are moving toward markets with more structural diversification.

Taiwan’s advantage is not merely about diversity for its own sake. The island’s semiconductor ecosystem includes TSMC, the world’s largest foundry, which serves as the manufacturing backbone for nearly every major AI chip design firm from Nvidia to AMD to Broadcom. It also includes packaging and testing leaders like ASE Technology and PacifiCare, as well as design houses like MediaTek and Realtek. An investor buying Taiwan equities is not betting on a single memory-chip cycle — they are betting on the entire AI hardware value chain, which is a far more durable proposition.

What this means for global chip demand

The cooling in Seoul is not necessarily a verdict on AI itself. It is a verdict on the valuation and concentration risk baked into the Korean market. But it does carry second-order meaning for global semiconductor cycles. When foreign investors — the very buyers who typically ride AI infrastructure build-outs — begin exiting a market on the grounds of demand uncertainty, it warrants attention from anyone tracking capex trends.

South Korea remains among 2026’s stronger-performing emerging markets on a full-year basis, supported by the massive gains in the first half and lingering expectations that AI infrastructure investment will continue. That context matters. The current pullback is a correction within a broader uptrend, not the beginning of a multi-year bear market in Korean tech. But corrections in concentrated, momentum-driven markets tend to be sharper and faster than in diversified ones, and the data so far fits that pattern.

The second-order effect is worth watching closely. Semiconductor demand signals tend to ripple through the global economy with a lag. If the Korean correction is telling us something real about the pace of AI infrastructure buildout — rather than just about valuation excess — then we should expect to see downstream effects in PCB orders, power-supply demand, and data-center construction timelines over the next two to three quarters. The memory-chip cycle is the leading indicator here. A pause or contraction in HBM demand would reverberate through the entire semiconductor equipment and materials chain.

There is also a geopolitical dimension to this rotation. Taiwan’s rise as the preferred AI-proxy market is partly a bet on institutional diversification away from single-name risk, but it is also a bet on the continued functioning of Taiwan’s semiconductor ecosystem amid escalating cross-strait tensions. That tension is not priced into the Taiex as fully as it should be. The market is rewarding diversification while underwriting a concentration of physical risk that no amount of portfolio theory can disguise.

Who wins, who loses

Taiwan wins in the near term. Its index has absorbed the rotation, and its broader supply-chain positioning gives it structural appeal as AI buyers diversify away from single-company risk. Global AI capex sponsors benefit too — a more diversified regional exposure reduces the chance that a single-cycle memory-chip slowdown derails the broader investment thesis. Portfolio managers who have rotated into Taiwan are better positioned for volatility, and they have the Diversification advantage that institutional risk frameworks increasingly demand.

Korean concentrated-position holders lose. Portfolio managers who loaded up on Samsung and SK Hynix during the first-half rally are now navigating a market where buyback support is fading, retail liquidity is retreating, and the very concentration that drove the outperformance is now the primary vulnerability. The $131 billion in foreign outflows this year is not a temporary wobble. It is a repricing.

The broader implication is that the AI equity story is no longer a single-country narrative. It never really was, but the market treated it like one for most of 2026. The rotation toward Taiwan and the steep withdrawal from Korea suggest institutional investors are finally pricing in that reality — and the next leg of the AI cycle will reward diversification far more than concentration. The question for the remainder of the year is whether the Korean market can stabilize on the back of earnings delivery rather than narrative momentum, or whether the outflows will accelerate as buyback floors disappear and retail participation continues to dry up. For now, the smart money is telling us exactly where it wants to be: on the island, not the peninsula.