【龔成教室】AI股有大崩盤前兆？散戶必睇：三星、SK海力士、美光，到底能笑到最後？｜HBM｜DRAM｜AI存儲股｜AI概念股｜龔成｜基本面分析 #HBM #SK Hynix #美光MU #三星

Core Theme

The video argues that AI-driven demand for HBM (High Bandwidth Memory) is a major tailwind for certain DRAM/AI memory “giants”, but it also warns that AI-memory stocks can see sharp volatility, driven by:

• Cyclical memory pricing
• Production/capacity swings
• Capital flow / “bubble mentality” in AI-related stocks

---

Companies / Tickers Mentioned

• Samsung (no ticker shown in subtitles)
• SK hynix (appears as variants like “SK can / SK Herix”; no ticker shown in subtitles)
• Micron (美光 MU) → MU
• “7709”: described as a leveraged product available in Hong Kong (likely a leveraged tracker/ETN; underlying not specified)

---

Key Portfolio Cautions & Recommendations (Explicit)

• Avoid leveraged products (example: 7709) because leverage + volatility can cause the instrument to lose value even if the underlying later returns to prior levels.
• Don’t chase “hot speculation” in AI stocks.
• Prefer waiting for market sentiment to normalize, and consider:
  • gradual entry
  • split orders
  • staggered approach

---

HBM Demand: Industry / Macro Explanation

Why HBM Matters

HBM is positioned as essential for AI compute because it delivers high-speed, high-volume data to GPUs, improving AI chip performance.

The speaker claims HBM demand has recently become an “explosive point” of growth.

Financial Link: AI Spending → Memory Demand

AI capex/spending by large tech ecosystems (examples mentioned: Google, Amazon, Microsoft) is treated as the downstream driver for memory demand.

However, the video warns that if AI platforms begin generating less cash flow or tighten spending, the memory supply-demand balance can change quickly.

---

Company-by-Company Notes (Fundamentals + Positioning)

1) SK hynix (SK海力士 / “SK can”)

• Positioning: described as a market leader for HBM.
• Revenue growth: stated to be more than 3x between 2021 and 2025.
• Utilization / profitability (Q2 2026):
  • operating utilization +54%
  • a figure implied to rise to ~76% (subtitle wording is unclear)
  • another 41% figure is referenced, but context is unclear due to subtitle errors
• Investment claim: if HBM demand stays tight and AI capex does not decline, SK hynix should be a major beneficiary.
• Risk emphasis: said to be most vulnerable if the cycle reverses due to production capacity exposure.

---

2) Micron (美光 MU)

• Positioning: characterized as having strong growth potential among the three.
• HBM revenue mix: HBM-based DRAM described as about “three-quarters” of company revenue (as stated).
• HBM market growth / share:
  • claims 2024 was “over the digits” growth
  • current share discussed around ~20% or more (base/year comparisons unclear due to subtitle errors)
• U.S. policy angle:
  • the video claims the U.S. government is focusing on domestic AI infrastructure
  • many large U.S. AI firms are said to build more in storage, benefiting Micron
• Profit/risk comparison:
  • Micron’s earnings level is described as not as high as Samsung’s, but its technical level is close, implying potentially higher growth
• Risk posture:
  • earnings are viewed as more diversified/resilient via revenue composition, though “earning capacity” may still be slightly lower than SK hynix/Samsung.

---

3) Samsung (三星)

• Financial tone: the video says profit has been growing and includes an internal comparison (subtitle wording is unclear).
• Business structure:
  • broader semiconductor exposure plus smartphones, automotive electronics, etc.
  • framed as reducing risk via diversification
  • however, the video claims Samsung’s HBM volume rate is generally considered weakest among the three
• Investment claim:
  • if Samsung’s HBM ramp improves, weaker initial volume positioning could create more upside
• U.S. strategic cooperation:
  • in July 2026, the video cites a $200 billion cooperation agreement involving major customers (examples: Google and Microsoft; subtitle text includes “MA”/possibly an Amazon typo)
  • the claim is improved customer base and visibility

---

Metrics / Numbers Explicitly Referenced

SK hynix

• Revenue growth: >3x (2021–2025)
• Q2 2026:
  • operating utilization +54%
  • utilization reaching ~76%
  • additional reference to 41% (context unclear)

HBM market (Micron-related statements)

• HBM market share: around ~20%+ in a recent period (comparison details unclear)

Strategic agreements

• July 2026:
  • SK group: cited $950 billion strategic cooperation
    • plus a $500 billion contract (names unclear in subtitles)
  • Samsung: cited $200 billion cooperation agreement

---

Risk Management / Investing Approach (Behavioral Checklist)

The video does not provide a formal mathematical framework, but it offers practical behavioral guidance:

• Time horizon: consider 5–10 years (or more) for “quality” AI industry stocks; short-term volatility is expected.
• Buy “quality”: focus on companies with strong fundamentals and technology leadership.
• Avoid leverage:
  • don’t buy leveraged Hong Kong products (example: 7709) because leverage can decay during volatility, even if the underlying later recovers.
• Entry timing:
  • don’t chase stocks when AI hype is “hot”
  • prefer staggered entry / split orders / gradual approach
• Track cycle risk:
  • monitor whether AI capex is still rising
  • remember memory is cyclical: if capitalization slows, memory pricing and earnings can fall quickly

---

Disclosures / Disclaimers

• The subtitles include a warning-style disclosure about leveraged speculation risk and cautions around product leverage.
• A generic “not financial advice” disclaimer is not clearly present in the provided subtitle text.

---

Presenters / Sources Mentioned

• Gong Cheng (龔成) — identified as the host/teacher in the title and subtitles (“龔成教室…龔成”).
• No other clear external analysts/sources are cited beyond general references such as:
  • U.S. government focus
  • major tech ecosystem firms (Google, Amazon, Microsoft) mentioned as demand drivers