Video summary

記憶體股票最後的機會，錯過，99%的人將要買在山頂。

Main summary

Key takeaways

Finance

Macro / Market Setup & Near-Term Risks

  • U.S. nonfarm payrolls appeared strong, but the speaker argues the headline is misleading:
    • Summer vacation hiring (e.g., hotels/entertainment) inflated job growth.
    • Early teacher hiring in August also boosted totals.
    • Information and finance employment fell substantially.

Key upcoming catalyst

  • U.S. CPI release
    • Focus on core CPI.
    • If core CPI unexpectedly rises, the speaker expects the U.S. may raise rates, pressuring markets.

Risk cautions (explicit)

  • Potential impacts from:
    • War
    • U.S. rate hikes
    • Japan rate hikes (noted as a factor that could influence U.S. tech, including tech stocks)
  • Mentions the possibility of a “black-swan crash.”
  • Practical guidance:
    • Don’t go all-in
    • Don’t pile all your family’s money
    • Don’t borrow to trade stocks

Trading / market reaction mentioned

  • After payrolls:
    • Philadelphia Semiconductor Index: +3.37%
    • SNDK: reportedly up up to +11%
    • SK Hynix: +7%
    • Other AI stocks also surged.

Investment Thesis: “Final Window” for Memory Stocks Tied to AGI/AI Acceleration

  • The speaker claims AI is entering an “AGI era,” citing:
    • OpenAI leadership
    • Jensen Huang (Nvidia)
  • Core argument: memory is the bottleneck/enablement for long-context reasoning.
    • New model capability (e.g., “Astra” / “Codex”) depends on vast memory to preserve and retrieve extensive prior information and tool outputs.
    • Belief: memory demand scales rapidly as reasoning generates large KV caches.
    • Eventually, reasoning-related data may total several terabytes per AGI-like task, potentially stored in NAND.

Bottom line: memory demand is expected to surge as systems move from “short context” toward longer, tool-using reasoning.

Method / Framework Presented (Step-by-Step Elements)

1) Use “fair value” tools to time entries/exits

  • Check valuations on InvestingPro:
    • Determine whether stocks are expensive vs. cheap.
  • Separate long-term vs short-term goals when using valuation metrics.
    • Warning: some stocks may already have future gains priced in.

2) Map the beneficiaries via bottom-up tech layers

The speaker outlines an infrastructure stack:

  • Tier 1: HBM
    • Examples: HBM3E, next HBM4/HBM5
  • Tier 2: DRAM / compute-adjacent memory (DRAM producers)
  • Tier 3: CXL + optical interconnects
    • Linking NAND pools
  • NAND layer: CXL-connected NAND demand rises as the ecosystem scales.

3) Risk management approach

  • Avoid “all-in” bets; accept possible drawdowns to participate in upside.
  • For most people:
    • DCA into index funds instead of stock picking (due to difficulty of multi-variable decision-making).

Quantitative Claims, Targets, and Timelines

Astra benchmark (as stated)

  • 99.9% on ARC-AGI-3

Nvidia systems used for Astra training (as stated)

  • 100,000 “Grace Blackwell NVL72” systems for training
  • Expected scaling to 400,000 units (4x) soon

Timeline emphasis

  • Now through end of September: expectation of continued news/partnership announcements for memory names (examples given include SNDK, Samsung, SK Hynix, Micron).
  • Technology ramp:
    • “More powerful NAND tech” in 2027
    • Mass production beginning 2028 (for SNDK/SK Hynix), referenced as HBF
  • Revisit timing:
    • Speaker suggests checking again “around this time next year” for a possible valuation reset / lowest prices.

Performance / valuation upside estimates (as quoted from InvestingPro)

  • Phison: ~45% upside
  • Nanya Technology: mentioned as a “keep an eye on” DRAM play (no explicit upside % provided in subtitles)
  • Innodisk (5289): ~22% upside
  • SNDK: described as a top pick (no explicit upside % stated in the subtitles for SNDK itself)

Company fundamentals / growth numbers (Innodisk)

  • Profit growth: +576.29% YoY
  • 6-month profit: “as much as the previous 20 years”
  • Revenue (first 8 months): NT$52.801 billion

Market-timing example caution

  • For high-flying “future vision” names (Tesla, SpaceX, Yageo):
    • Speaker warns you may need to accept roughly ~20% short-term decline risk to capture analysts’ expected upside.

Specific Stocks / Tickers / Sectors and How They Fit the Thesis

Memory & semiconductor (primary focus)

  • SNDK (SanDisk / Western Digital entity referenced in subtitles; treated as a key NAND/CXL/optical interconnect beneficiary)
  • SK Hynix
  • Samsung
  • Micron
  • Kioxia (Japan; NAND; supports/pushes into CXL per subtitles)
  • Marvell
  • ALAB (unclear acronym in subtitles)
  • Broadcom (AVGO) (supported as optical interconnect/platform beneficiary for the CXL ecosystem)

CXL-related “major memory companies” list (as stated)

  • SK Hynix, Samsung, SNDK, Micron, Kioxia

Other cited memory/ecosystem plays

  • Innodisk — ticker 5289 (CXL modules; earnings surge cited)
  • Nanya Technology (DRAM maker)
  • Phison (NAND controller/storage ecosystem; tied to CXL/NAND module demand; upside estimate mentioned)

Other “caution/tone” references (not memory-focused)

  • Tesla
  • SpaceX
  • Yageo
    • Mentioned in the context of potentially tolerating ~20% short-term decline risk.

Explicit Recommendations / Calls to Action

  • Bullish stance: long-term bullish on U.S. and Taiwan tech/memory stocks into next year, especially around/after midterm elections.
  • Entry timing: use InvestingPro fair value and valuation levels to find “final opportunity” entries.
  • Positioning/risk:
    • Never go all-in
    • Avoid borrowing to trade stocks
  • For most individuals:
    • DCA into index funds rather than direct stock picking

Disclosures / Disclaimers

The speaker states:

  • “This video is only my personal investment journal.”
  • “These are purely my personal views.”
  • Claims they “openly discuss every stock I like,” and viewers should do their own research.
  • Mentions financial product promotion/discount mechanics for InvestingPro / PRO+ (via QR code/pinned link/discount offer).
  • No explicit “financial advice” wording appears, but the personalization + do your own research disclaimer is included.

Presenters / Sources Mentioned

  • Jensen Huang (Nvidia)
  • OpenAI (subtitles reference “Welcome to the AGI era,” attributed to OpenAI leadership)
  • Micron (roadmap referenced)
  • ARC-AGI-3 benchmark (third-party benchmark used for “Astra” claims)
  • Investing.com / InvestingPro (valuation guidance tool/platform)

Additional ecosystem companies cited:

  • SK Hynix, Samsung, Micron, Kioxia, Marvell, Broadcom (AVGO), Phison, Nanya Technology, Innodisk (5289), plus ALAB (as written in subtitles)

Original video