Korea's Crash: Is the AI
Boom Coming to an End?
Oct 26, 2026
8 min read
Global Markets
Executive Summary
Korea's Crash Is Not the End of AI
The KOSPI's nearly 40% decline reflects a sharp correction in concentrated, leveraged semiconductor positions rather than a collapse in underlying AI demand.
Strong Fundamentals, Falling Stocks
Samsung's record profits, surging chip exports and resilient AI orders highlight a growing disconnect between business performance and valuations in the cyclical memory-chip industry.
Debt Is the Real AI Risk
AI infrastructure spending continues to expand, but increasing reliance on borrowing amid elevated interest rates raises concerns about the sustainability of the investment cycle.
A Potential Opportunity for India
An unwinding of crowded North Asian AI trades could redirect global capital towards Indian equities. Quality large caps and IT services may benefit, while leveraged and AI-capex-driven businesses warrant caution.
Indicator
Reading
Read-through
KOSPI
5,533.77 (22 Oct)
39.3% below the 9,114.55 close of 22 June
KOSPI, last 90 days
−17.3% (from 6,690.62, 24 Jul)
Two crashes, one rebound: the '40%' spans four months
Samsung + SK Hynix weight
~50% of KOSPI market cap
An index that is really a memory-chip fund
SK Hynix share price
−57% from June peak
~2.7x 2027 earnings (BofA); prior upcycles 4–7x
Samsung Q3 operating profit
₩107.4tn (+782% YoY)
A record quarter, delivered into a falling share price
Korea chip exports, 1–20 Oct
US$27.5bn (+222.5% YoY)
The physical demand has not cracked
Margin loans (13 Oct)
₩34.25tn
Forced selling of ₩286bn in four sessions, 13–16 Oct
Foreign net selling of Korea
>₩200tn in 2026
Foreigners sold the rally to leveraged locals
Hyperscaler capex (FY26)
>US$690bn (FactSet)
Still rising; debt now funds ~32% of increments
India FPI equity flows, 2026
−₹3.12 lakh crore (to ~10 Oct)
Money that left India partly chased North Asian AI
Source: Ametra Research
01
South Korea's KOSPI lost nearly 40% from its June peak in a volatile four-month period, interrupted by a dramatic recovery that ultimately failed to hold.
What Drove the Two Sell-Offs?
July: The First Shock
October: The Second Wave
After a sharp rebound fuelled by renewed AI optimism, selling pressure returned, pushing the index back towards its July lows.
“
The Pattern Matters More Than the Fall
A market that crashes, rebounds sharply and then returns to its lows suggests more than a simple reassessment of AI demand. Concentrated positions, leverage and repeated selling pressure have amplified the correction.
02
Korea's crash wasn't simply about its exposure to AI. It was about how that exposure was structured — three vulnerabilities amplified one another.
Concentration
~50%
Leverage
₩34.25T
Cyclical Memory Chips
“
The Crash Was Amplified by Market Structure, Not Just AI Concerns
Foreign investors sold more than ₩200 trillion of Korean shares during 2026, while leveraged domestic investors absorbed much of the selling. Concentrated exposure, margin debt and cyclical earnings transformed a correction into a far deeper market decline.
03
Korea's semiconductor industry continues to report strong earnings and exports, even as its leading chipmakers face a sharp market correction. The disconnect raises an important question: why are share prices falling when business performance remains strong?
The Earnings–Market Disconnect
● Operating profit, ₩ tillion
Share Price declined by 39% from 17 Jun 2026 to 20 Oct 2026 despite record profits
Why Are Share Prices Falling?
Memory-chip stocks are valued on expectations of future earnings, not simply today's profits. Despite record results, investors are questioning whether exceptionally high memory-chip margins can be sustained through the next cycle. Samsung and SK Hynix trade at roughly 3.0x and 2.7x estimated 2027 earnings, respectively, reflecting concerns that current profitability may be near a cyclical high.
“
The Market Isn't Questioning AI Demand. It's Questioning Peak Earnings.
Record profits and falling stock prices can coexist when investors expect earnings to decline from unusually high levels. The central question is whether today's exceptional memory-chip profitability can last, not whether demand for AI infrastructure still exists.
04
The Real Fault Line –Who Pays for the Boom
AI infrastructure spending continues to accelerate, but a growing share is being funded through borrowing. The concern is no longer whether AI demand exists, but whether future revenues can justify the scale and rising cost of investment.
AI Spending Is Rising. So Is Its Dependence on Debt.
● Capex, US$ bn
New debt as a % of capex has increased from 9% to 32% from FY24 to LTM Jun 2026
Why Financing Is Becoming the Bigger Risk
Higher Borrowing Costs
Credit Quality Under Pressure
Infrastructure Bottlenecks
“
The Real Risk Isn't AI Demand. It's the Cost of Sustaining It.
AI investment continues to expand, but the shift from cash-funded growth towards debt-financed expansion makes the cycle increasingly sensitive to interest rates and credit conditions. The next test will be whether AI revenues can justify the enormous capital commitments being made today.
05
+41%
TSMC Revenue Growth
Q3 2026 · YoY
-1.7%
Taiwan TAIEX
First half of October
2-3x
Korean Memory Valuations
Estimated earnings multiples
The Dot-Com Bubble vs. Today's AI Cycle
Factor
2000 · Dot-Com
2026 · AI Cycle
Valuations
Cisco exceeded 150x forward earnings at its peak.
Korean memory makers trade around 2–3x estimated earnings.
Demand Evidence
The comparison centres on the historic valuation excess.
TSMC reports robust AI orders and accelerating revenue growth.
Central Question
Could valuations justify expectations?
Can peak earnings and AI investment spending remain sustainable?
Two Risks That Still Matter
Peak Earnings Risk
2027 Capital Expenditure
Upcoming hyperscaler spending budgets will determine whether the current pace of AI infrastructure investment can continue.
“
The AI Demand Story Remains Intact. The Investment Cycle Still Faces a Test.
Strong semiconductor orders and relatively low memory-stock valuations distinguish today's AI cycle from the dot-com era. However, low multiples provide limited protection if earnings decline, and the next round of hyperscaler spending decisions will be critical in determining whether the boom remains sustainable.
06
₹3.12L Cr
FPI Equity Outflows
India · 2026 to 16 October
10.65%
India's MSCI EM Weight
September 2026
+11.7%
Nifty IT Weekly Return
Week ending 16 October
Where the Opportunities and Risks Lie
Segment
Ametra's Perspective
Quality Large Caps
Could benefit as global capital rotates away from crowded North Asian AI trades.
Indian IT Services
Potential beneficiaries of improving relative sentiment; look for confirmation of sustained flows.
AI Infrastructure
Data centres, power equipment and cables remain exposed to a global AI capex cycle facing valuation scrutiny.
Leveraged & Momentum Trades
Warrant caution as concentrated positions and leverage can amplify market corrections.
“
The Unwinding AI Trade Could Put India Back on Investors' Radar
India's relatively limited AI-hardware exposure and already-corrected large-cap valuations improve its position as global investors reassess concentrated AI bets. However, elevated US yields may delay a sustained reversal in foreign flows. The opportunity lies in selective exposure to quality businesses, not indiscriminate buying of AI-linked or momentum stocks.
07
Stance
Where
Rationale
Favour
Indian quality large caps
Low AI-hardware exposure; best placed if flows rotate out of North Asia
Add-on confirmation
Indian IT services
Rebounding as the AI trade unwinds; add on dips if flows turn
Selective
Domestic AI-capex plays (data centres, power, cables)
Priced for a capex cycle now being repriced
Watch
Korean memory makers
2–3x P/E only cheap if earnings hold; wait for 2027 capex guidance
Caution
Concentrated global AI and semiconductor funds
Valuation and financing risk, not demand risk
Underweight / trim
Leveraged and momentum positions
Korea shows how leverage turns a correction into a crash
Watch
Brent, INR, October CPI, RBI's December meeting
The tripwires that would turn a plateau into a second leg
“
08
Ametra’s Read
We believe Korea's sharp market correction signals the end of the easy AI trade, not the end of the AI boom. Strong semiconductor profits, chip exports and AI demand suggest that the underlying growth story remains intact, while excessive leverage, stretched expectations and rising financing costs have exposed vulnerabilities in the investment cycle. Our focus remains on balance-sheet strength, sustainable earnings and valuations rather than headline growth or seemingly cheap multiples. We remain cautious on leveraged AI and memory-chip investments, with AI-linked credit spreads and hyperscaler capital expenditure as key indicators to monitor. For Indian investors, a potential rotation away from crowded North Asian AI trades could create selective opportunities in quality large-cap equities and IT services, while domestic AI infrastructure plays warrant greater scrutiny.
