South Korea’s AI-linked chip stocks have taken another sharp hit, and the instinct is to read that as evidence the AI boom itself is losing steam.
The more accurate read, according to the numbers underneath the selloff, is narrower than that: semiconductor exports remain robust, and what’s actually happening looks like investors taking profits off a trade that ran further and faster than almost anything else in global markets over the past two years, not a sign that AI demand is cracking.
Why this looks like profit-taking, not a demand problem
The timing undercuts the “AI is slowing” narrative on its own.
South Korea has launched a state-backed investment program worth about $518 billion. Samsung Electronics, SK Hynix, and their suppliers plan to build new chip fabrication plants in the country’s southwest.
The initiative supports President Lee Jae Myung’s goal of strengthening South Korea’s leadership in semiconductors, physical AI, and data centers.
That’s not the posture of an industry bracing for a demand collapse. What’s moving instead is valuation.
Korean and Taiwanese chipmakers have functioned as close to pure plays on the AI infrastructure buildout, and pure plays are exactly what gets sold first once investors start questioning whether near-term earnings can keep justifying prices that assumed years of uninterrupted growth.
Nvidia and Alphabet have followed a similar pattern in recent months. Both companies pulled back after reaching record highs. The decline reflected a valuation reset rather than weaker business fundamentals. Their long-term growth story remains intact.
Why India keeps coming up as the rotation candidate
Jefferies strategist Chris Wood has been making a specific case for months that he calls the “reverse AI trade”: once investors start questioning returns on the roughly $620 billion hyperscalers are spending annually on AI infrastructure, capital concentrated in chip-heavy markets like Taiwan and South Korea could rotate toward markets with less direct AI hardware exposure but stronger domestic growth engines, India chief among them.
The irony is that India’s lack of exposure to the AI hardware supercycle, the same trait that made it underperform through 2025 and into 2026, is precisely what makes it the candidate now.
Foreign investors pulled nearly $33 billion out of Indian equities since late 2024 chasing the Taiwan and Korea trade, and India’s weighting in the MSCI Emerging Markets Index, at roughly 12%, now sits below TSMC’s weighting alone.
What’s kept Indian markets from cracking under that outflow is domestic money: monthly SIP inflows running around $3.4 billion have absorbed much of the foreign selling, while corporate earnings have grown in the mid-to-high teens percentage range, a fundamentally different growth driver than a semiconductor cycle tied almost entirely to AI capex decisions made in Silicon Valley boardrooms.
None of this is a guarantee India actually captures a large rotation if the Korea correction deepens, it’s a thesis that depends on the AI trade cooling further rather than simply catching its breath.
The number worth watching from here isn’t a single day’s Kospi move, it’s whether Samsung and SK Hynix’s actual order books soften in the coming quarters. That’s the line between a healthy pullback in an intact boom and the start of the unwind Wood has been describing.
Source: Moneycontrol, "Has the AI Semiconductor Trade Peaked? Korea's Rout Could Be the First Sign of a Global Rotation"
