Summary of "반도체 지금 꽉 쥐세요 (feat.삼성액티브자산운용 김효식 팀장)"

Summary of Key Arguments and Analysis

Shift in Valuation Logic: From PBR to PER

The speakers argue that Korea’s listed semiconductor and power-equipment companies have seen a dramatic re-rating.
The re-rating is attributed to a move away from:
- Asset-value-based valuation (PBR) toward
- Earnings-based valuation (PER)
Core claim: when earnings become more stable and predictable, investors may stop treating these firms as purely cycle-driven and instead value them as durable profit businesses.

Power Equipment (Transformers) as the “Case Study”

The video uses the Korean transformer industry as a template, suggesting similar dynamics are beginning to appear in semiconductors.

Key reasons cited for the transformer rally:

AI-driven electricity demand
- Data centers are described as consuming over 50% of incremental power demand (per the referenced chart).
Power shortages and U.S. infrastructure bottlenecks
- Connection-to-commercial-operation timelines reportedly increased from ~2.3 years (around 2000) to about 5–4.8 years (2024).
Long-term supply contracts emerging for U.S./Europe customers
- Starting around 2023, the video claims contracts have begun to appear, including:
  - Multi-year agreements
  - Advance payments (e.g., 10% down / interim payment structures)
  - Escalation clauses tied to cost increases

Major thesis:

Long-term contracts can shift the transformer business from a highly volatile cycle to a model with:
- steadier volumes
- more stable margins
That stability, in turn, supports the valuation shift from PBR to PER.

Supporting evidence mentioned:

Market-cap growth is presented as confirmation, with Hyundai Electric, Hyosung Heavy Industries, and LS Electric rising from:
- ~2.3 trillion KRW (early 2019)
- to ~104 trillion KRW (Apr 23, 2026)
This is framed as roughly a 46x increase, alongside rising KOSPI weight from ~0.2% to ~2.0%.

Translating the Same Mechanism to Memory Semiconductors

For memory, the argument is not only that AI increases demand, but that AI is changing how demand behaves.

The video describes:

Bottlenecks emerging across the memory stack, not just HBM
- HBM → NAND/SSD/HDD-class needs via NAND
- Plus broader DRAM usage growth
A shift in AI workloads:
- From “training-only” to inference / agentic AI
- This increases memory needs for larger context retention, tightening supply across the system

Long-Term Contracts as the Semiconductor “Trigger”

The most important parallel to transformers is claimed to be the spread of long-term supply contracting for memory, extending toward 2027–2028, with especially strong focus on locking in the final 2–3 years.

Claimed effects on volatility:

Without long contracts (a spot/short-cycle market), companies can suffer severe deficits in downcycles.
With longer-term contracts:
- Upward potential may become less explosive
- Downside volatility should weaken
- The overall profile becomes more structural growth than pure cyclicality

The video also suggests that once contracts become “binding” (e.g., via disclosed advance payments), analysts may be forced to reconsider valuation frameworks.

Valuation Implication for Samsung/SK Hynix and Index Levels

The speaker claims that for Korea:

Forward consensus P/E for Samsung Electronics and SK Hynix in 2026–2027 is under ~5x
By contrast, the “normal” historical context for KOSPI is framed as closer to ~10x

If earnings sustainability improves and valuation shifts from PBR to PER, the video suggests:

A possible re-rating
- Framed cautiously as potentially “2x upside” due to the valuation-logic change
A scenario where KOSPI could reach 8,000
- Primarily based on Samsung + SK Hynix’s combined earnings/valuation shift
- With the caveat that macro shocks could still derail the outcome

Caution Against Chasing Recent Stock Gains

Despite optimism, the video repeatedly warns against assuming:

“profits are fully taken” merely because stocks have already risen sharply.

Instead, it emphasizes tracking:

contract disclosures
changes in valuation methodology
not just spot demand peaks

Technological Progression as Continuing Support

The speakers close by arguing AI-related memory evolution is still ongoing, including:

iterative HBM upgrades
DRAM variant development
NAND roadmap ideas

They emphasize memory is not a simple commodity cycle; it has:

technology-driven pricing
evolving unit economics shaped by design changes to manage latency/bandwidth constraints

Presenters / Contributors

Choi — host/director
Kim Yo-sik — Team Leader, Samsung Active Asset Management
Kim Hee-sung — mentioned as a contributor/enthusiast in the introduction (also referenced as a “power equipment enthusiast”)
(Stated elsewhere) “Lee Kyung-jae (Energy, SK)” and other analysts (named as examples), but not presented as video participants in the discussion panel.