Ideas
AI semiconductor momentum persists despite bubble noise.
The AI speed-control debate is unrealistic because companies cannot credibly coordinate to slow development (prisoner's dilemma and free-rider problems, as seen with OPEC quota cheating), and they cannot fully open their technology books. Bubble concerns are amplified by anxious sentiment when stocks are rangebound, but industry leaders Jensen Huang and Chey Tae-won indicate supply will increase and shortages will persist through 2028-2030. The latest evolved bubble theory is the share-loss narrative (Samsung taking SK Hynix's HBM share, CXMT entering DRAM), but the total market pie is growing, so both can grow. Operating margin concerns remain a watch point for Q3, especially with FX effects, but have not materialized. Semiconductors are the constant core of the AI cycle.
Power infrastructure is AI's critical bottleneck.
As AI token/data accumulation drives massive data center buildout, power is becoming the critical bottleneck. US data centers are being halted due to lack of energy, and Trump's executive order tells operators to self-generate power. SMR/nuclear takes too long, so near-term solutions include fuel cells, transformers, transmission/distribution, and power cables. Power infrastructure is the vessel that contains AI. This week's market rotation (Wed: transformers, Thu: ship engines, Fri: power cables) shows the theme emerging. He recommends gradually accumulating power infrastructure-related stocks as the AI cycle evolves; long-term (5-10 years), nuclear/SMR will also bloom as data centers become power-hungry.
Telecom/6G infrastructure needed for AI expansion.
Data centers need not only power but also communication infrastructure to operate. 5G is insufficient; 6G with at least 4x the speed is required for autonomous driving data and humanoid robots. Optical communication and general telecom infrastructure are needed. He flags telecom/communication infrastructure as a key focus for next year alongside semiconductors and power infrastructure, recommending gradual accumulation of related stocks.
Avoid lagging non-AI sectors; focus AI infra.
Familiar sectors such as pharma/bio, securities, cosmetics, and robots have not participated in the AI-driven rally—pharma/bio and securities have left the market, cosmetics recently plunged on FX, and robots are in a fog with potential for intensifying competition. He advises discarding fixed ideas about these lagging sectors and focusing on the AI industry cycle evolution (semiconductors, power, telecom). Batteries are more mixed (somewhat included via ESS potential), but not his main focus.
Buy Samsung Electronics; watch foreign flows.
If he had 10 million won right now, he would buy Samsung Electronics. Near-term catalysts: 30 trillion won dividend with entitlement date around Sept 28 (4 trading days left); if foreigners don't buy in to receive the dividend, it signals they have no interest; if they do, it's meaningful. Also, Samsung and Hynix buybacks (15 trillion + 40 trillion = 55 trillion won) end mid-October; historical pattern shows foreigners sell during buybacks then the stock rises after buybacks end when the buyer disappears. Both scenarios (foreigners entering for dividend, supply dynamics post-buyback) are watch points. Caveat: for a 10-year horizon, he would not buy Samsung due to Chinese memory competition (CXMT) and intensifying competition; semiconductors are a 5-year confidence, not 10-year.
Buy Hyosung Heavy Industries for 10 years.
If investing for his son over 10 years, he would buy Hyosung Heavy Industries. It is a power infrastructure play with about 12% global market share; despite competition from GE, GS Hitachi, etc., the spillover effect from the growing power market is huge. Unlike semiconductors (Samsung) where 10-year visibility is clouded by Chinese memory competition (CXMT) and intensifying competition, power infrastructure has clear 5-10 year growth potential as data centers, SMR, and nuclear drive massive electricity demand. He is confident about semiconductors for 5 years but not 10; power infrastructure is his 10-year idea.
This 3PRO TV (삼프로TV) video, published September 18, 2026,
features Cha Young-joo
discussing SMH, Power infrastructure, Telecom/communication infrastructure, XBI, Korean securities sector, KORU, Korean robotics sector, 005930.KS, 298040.KS.
6 trade ideas extracted by AI with direction and confidence scoring.
Speakers:
Cha Young-joo
· Tickers:
SMH,
Power infrastructure,
Telecom/communication infrastructure,
XBI,
Korean securities sector,
KORU,
Korean robotics sector,
005930.KS,
298040.KS
