South Korea's Kospi Falls Sharply as U.S. AI-Chip Selloff Spreads to Samsung and SK Hynix

Kospi's Sharpest Week in Years

South Korea's benchmark Kospi index fell sharply across multiple trading sessions in July 2026, led by declines in the country's two dominant chipmakers, Samsung Electronics and SK Hynix. The single-day drops ranged from about 4.6% to 7.89%, with the steepest coming on July 2, when the index tumbled 7.89%, and a second severe session on July 17, when it fell 6.37%^[1][11]. Trading in Seoul grew volatile enough at points to trigger exchange "sidecar" mechanisms, which briefly pause program selling to prevent panic from feeding on itself^[14].

The trigger originated across the Pacific. The week before, a selloff in U.S. semiconductor stocks had erased more than $1 trillion in combined market value, with the Philadelphia Semiconductor Index falling roughly 10%, Intel shares dropping about 21% over several trading days, and Micron losing roughly $38 billion in market value in a single session^[3][4]. Nvidia also sold off. As those losses rippled outward, Samsung and SK Hynix — both deeply tied to the global AI-chip supply chain — absorbed the shock, and Korea's broader market absorbed it with them^[1].

By the following week, U.S. stock futures were attempting a partial rebound, in a stretch that had already produced both steep drops and quick recoveries^[7][14]. The whiplash captured the essence of the moment: investors were not disputing that AI infrastructure spending is enormous, but arguing fiercely over whether it is sustainable.

What Both Sides Concede

Stripped of interpretation, the numbers are not in dispute. The four largest U.S. cloud providers — Alphabet, Amazon, Meta and Microsoft — have guided to roughly $725 billion in combined 2026 capital spending, the great majority of it aimed at AI infrastructure^[9]. Goldman Sachs, meanwhile, projects that big-tech AI capital spending will exceed $1 trillion in 2027, a forecast that assumes the buildout keeps accelerating rather than stalling^[15].

It is also uncontested that Samsung and SK Hynix together now make up roughly half of the Kospi's total weight, up from about a quarter at the end of 2025^[1]. And even as share prices swung wildly, many Wall Street analysts kept their 12-month price targets on chipmakers such as Nvidia and Micron in place, describing the selloff as a valuation reset rather than a sign that chip earnings themselves were deteriorating^[4].

The Pressure Underneath

Three structural realities sit beneath the daily headlines, and none of them resolves the argument on its own. The first is concentration risk: because Samsung and SK Hynix carry roughly half the Kospi's weight, the entirety of South Korea's stock market now rises and falls on global sentiment about AI memory chips, regardless of what either company's actual order book looks like^[1].

The second is a capex-revenue gap. Hyperscalers have committed roughly $725 billion for 2026, but the AI revenue needed to justify that spending has been arriving more slowly than the money going out the door, creating a structural anxiety that exists independent of any single trading session^[9][5]. The third is competitive repricing: custom AI chips from OpenAI, Amazon and Cerebras began shipping in volume in late June and early July 2026, a genuine shift that narrows Nvidia's historical dominance of AI compute and reshapes where capital flows next, regardless of which broader narrative about the selloff turns out to be right^[3].

Taken together, the physical facts of the buildout are not really contested: enormous real capital is flowing into data centers and chips, demand for the high-bandwidth memory used in AI servers remains strong, and chipmaker earnings remain solid^[3][10][4]. What is genuinely uncertain is the return on that spending and the timing of when AI revenue catches up to the outlays^[5].

How Each Side Sees It

Skeptics of the AI buildout point to a timeline mismatch: hyperscalers are committing hundreds of billions of dollars now, while the AI revenue to pay for it has not clearly materialized, and cheaper Chinese and open-source models threaten to compress prices further. This camp draws comparisons to the market conditions of June 2000, just before the dot-com crash, and warns that if data-center budgets are cut, corporate investment in the broader economy — which they say is barely positive without IT spending — could stall^[5][6]. Their incentive is to price in risk before a possible correction and protect capital; each downdraft in the market tends to validate their positioning, even as a sustained rebound would leave them having missed further gains^[6].

AI bulls, including chip industry figures, hyperscalers and many Wall Street analysts, argue the selloff is better understood as a "mid-cycle reset" of overstretched prices rather than evidence that AI demand is fading. They point to strong chipmaker earnings, soaring demand for HBM chips used in AI servers, and Goldman Sachs' own forecast of capex rising past $1 trillion by 2027 as evidence that the underlying business remains healthy. They also note that Apple, which spends comparatively less on capex and instead partners for AI capabilities, has been the best-performing megacap stock this year — evidence, in their reading, that markets are repricing risk rather than abandoning AI altogether^[4][10][3]. Their incentive is to defend valuations and keep capital flowing into a buildout that drives their own revenue, and Samsung, SK Hynix, Micron and Nvidia are directly exposed to whichever way sentiment breaks, even as their AI order books stay full^[1][3].

South Korean commentators frame the story differently again, stressing that part of the pain is a home-grown structural risk. With Samsung and SK Hynix now at roughly half the Kospi's weight, the national index has effectively become a leveraged bet on AI memory chips, and local triggers — including a Bank of Korea interest-rate decision and forced selling by margin investors — amplified the U.S.-driven move into what one outlet called a "roller coaster"^[11][13]. Retail and pension investors heavily exposed to the two chipmakers faced large paper losses and forced liquidations during the sharpest drops, in an export-dependent economy tightly bound to global memory-chip demand^[13][14]. A related thread, more focused on competitive dynamics than macro sentiment, holds that Nvidia's near-monopoly on AI compute looks genuinely more contested than it has in years, as custom silicon from rivals begins shipping at scale — a shift that would reshape where AI-capex dollars flow no matter how the broader bubble-or-reset debate resolves^[3].

How the Coverage Split

Coverage of the selloff diverged less over the facts than over which facts to foreground. U.S. business outlets closer to the market center, such as CNBC and Bloomberg, stuck largely to mechanical, cause-and-effect reporting — percentages, index levels and phrases like "AI concerns weigh" — without taking a position on whether the spending is justified^[1][2]. Forbes, reflecting a more market-bullish lean, emphasized analysts holding their price targets and framed the episode as a "mid-cycle reset," a reading that reassures investors the underlying fundamentals remain intact^[4].

Outlets further to the left of the U.S. spectrum leaned toward the skeptical read: CBS News framed the story around investors wanting "proof" that AI spending would pay off, while NPR posed the pointed question "Is AI 'one big bubble'?" and invoked dot-com-era comparisons, priming readers toward doubt even while phrasing it as a question^[5][6].

Non-Western outlets largely sidestepped the American framing war altogether. Korea's KED Global reached for vivid domestic imagery — turning the Kospi into a "roller coaster" — to center concentration risk and forced selling, while India's Sunday Guardian took an explainer-style approach aimed at retail investors, heavy on percentage moves and trading-halt mechanics but largely neutral on the underlying AI debate^[11][13].