Video summary

AI Stocks will DIE or become Multi-baggers? (Detailed Analysis) | Akshat Shrivastava

Main summary

Key takeaways

Finance

Finance-Focused Summary (AI Stocks Correction Thesis)

Market Context / Performance Snapshots

  • KOSPI (South Korea index): down about 45% in one month (used to illustrate aggressive drawdowns).
  • QQQ (Nasdaq-100 ETF): down about 11.63% from peak (used to argue the correction isn’t “the end” everywhere).
  • Micron (MU): down about 40–41% from peak.
    • Despite the ~-40% drawdown, the speaker emphasizes Micron is still up ~500% from earlier levels, framing this as a valuation/price correction after a massive run-up, not a business collapse.

Core Claim / Framework: “AI Has 3 Buckets”

The speaker argues AI is not a single trade; it’s segmented into three layers, and only one layer is currently correcting.

  1. Hyperscalers (Layer 1) — “Doing Fine”

    • Examples: Nvidia (NVDA), Microsoft (MSFT), Meta (META), Alphabet/Google (GOOGL/GOOG).
    • Thesis: operating profits / revenue growth remain intact; drawdowns like ~17–20% are treated as normal volatility after large prior rallies.

  2. AI Built-Out / Picks-and-Shovels Supply Chain (Layer 2) — “Where the Major Problem Is”

    • Examples mentioned: Micron (MU), SanDisk (brand mentioned; no separate ticker quoted), SK Hynix (ticker not explicitly stated), Oracle (ORCL), Nebius (not clearly identifiable as a public-market ticker in the subtitles), plus related memory and compute-enabling infrastructure.
    • Thesis: this segment ran up extremely fast (Micron cited as ~530–600%); the current correction (about ~40–65% on examples) is framed as a valuation reset.

  3. Consumer-Facing AI Apps (Layer 3) — “Also Intact”

    • Examples mentioned: Meta (also placed in this bucket), Adobe (ADBE), Duolingo (DUOL), plus Google (as both hyperscaler and consumer-app exposure).

Key Takeaway (repeated): The AI bubble is not “burst” because only Layer 2 is correcting. Layers 1 and 3 remain intact. The correction is characterized as valuation correction, not proof that AI businesses are “going rogue.”

Macro / Earnings & Guidance Signals Used

Capex Cycle Remains Central

  • Hyperscalers are planning higher capex, with capex peaking around 2027.
  • Continued buildout is framed as supportive of Layer 2 suppliers, assuming entry valuations are reasonable.

Company Results / Interpretation (High-Level)

  • Google/Alphabet: “earnings and revenues have gone up” (presented as positive).
  • Meta: revenues up, but earnings slower → stock down ~9–10%.
    • Management suggests they won’t cut capex, because compute capacity needs will increase dramatically.
  • Microsoft: revenues/earnings up; guidance framed as “more rational” spending, while the speaker still argues capex rises through 2026–2027.

Meta CFO / Management Quote (Disclosure-Free Excerpt)

  • The CFO’s stance (as relayed): they don’t want to fall behind in compute demand; they accept a short-term profit hit to increase spend.

Valuation & Earnings-to-Price Reasoning

“Operating Profits Doubling” as a Buy Argument (Hyperscalers)

  • The speaker claims operating profits double about every ~5 years for Meta and Microsoft.
  • Implication: if earnings growth stays on track, temporary multiple compression doesn’t necessarily invalidate the business.

Explicit PE Multiple Contrast (Examples)

  • Meta: described as having a “sensible” PE, not comparable to speculative very high-PE growth story stocks (examples referenced include Zomato and Lenskart, mentioned as PE 120–200+ style comparisons—tickers not provided in subtitles).

Recommendations / Portfolio Actions (Explicit)

Timing / “Don’t Chase”

  • Guidance: don’t chase a falling/running asset; instead, let corrections consolidate into a buying zone.
  • Caution: buying after parabolic peaks is framed as poor entry timing (“you deserve to lose money”).

Capital Allocation Style: Profit Booking + Rotation

  • Advocate a rotation/trading approach rather than only buy-and-hold:
    • Book profits during strong run-ups and rotate capital.
    • Example theme: cybersecurity names (Palo Alto (PANW), CrowdStrike (CRWD)) ran up >100% in a few months, then corrected.

Specific Action on Meta

  • Plan: double down on Meta.
    • Current position: ~10% / ~10.5% of portfolio.
    • Target: increase to ~15%, “depending on how aggravated the fall is.”
    • Timing: “start doing this from today.”

Micron Stance

  • Micron is not rejected. The speaker expects a re-buy later, reiterating that the -40% drawdown is an opportunity after consolidation, not something to chase immediately.

Key Numbers & Performance Metrics Mentioned

  • KOSPI: -45% in one month.
  • QQQ: -11.63% from peak.
  • Micron (MU):
    • -40% to -41% drawdown,
    • still ~+500% overall from a prior period,
    • rally described as ~530–600% in about 1–1.5 years.
  • Speaker’s disclosed portfolio drawdown:
    • Peak: $4.0M
    • Now: $3.846M
    • Drawdown: about $150k, described as roughly ~3%
  • Meta stock move (earnings reaction): ~9–10% correction.
  • Hyperscaler drawdowns (examples):
    • Nvidia (NVDA): ~17–18% from peak (as framed)
    • General statement: ~20% drawdowns can happen in hyperscalers

Disclosures / Disclaimers

  • The subtitles include a claim that the speaker’s guidance is based on “real investing experience” and that he manages money/hedge fund.
  • No explicit “not financial advice” disclaimer appears in the provided subtitles (so none can be confirmed from the text).

Instruments / Tickers Explicitly Mentioned

  • QQQ
  • Micron (MU)
  • Nvidia (NVDA)
  • Microsoft (MSFT)
  • Meta (META)
  • Alphabet / Google (GOOGL/GOOG)
  • ASML (ticker: ASML)
  • SK Hynix (ticker not stated)
  • Oracle (ORCL)
  • ServiceNow (NOW)
  • Adobe (ADBE)
  • Duolingo (DUOL)
  • Palo Alto (PANW)
  • CrowdStrike (CRWD)
  • Zomato (mentioned as a high-PE example; ticker not stated)
  • Lenskart (mentioned as a high-PE example; ticker not stated)
  • KOSPI (South Korea index)
  • SanDisk (brand mentioned; no ticker stated)
  • Nebius (named; ticker not stated)
  • SpaceX (mentioned but excluded due to IPO comparability)

Step-by-Step / Methodology Elements Shared

  • Segment AI into 3 layers:
    1. Hyperscalers (data centers/builders)
    2. AI built-out (infrastructure supply chain: memory/compute rentals, software enablers)
    3. Consumer-facing AI apps/products
  • Evaluate corrections by layer:
    • If Layer 2 corrects while Layers 1 & 3 remain supported → interpret as valuation consolidation, not necessarily “bubble burst.”
  • Validate with capex guidance:
    • Check hyperscalers’ statements that capex continues to rise, peaking around 2027.
  • Buy discipline:
    • Prefer buying after correction consolidates (“buy in zone”).
    • Avoid chasing.
  • Portfolio management approach:
    • Use profit booking and rotation during strong run-ups (especially in high-beta areas), then reallocate.

Presenters / Sources

  • Presenter: Akshat Shrivastava
  • Named market participant mentioned (as a tweet source): Vikas Khetani (referenced as “celebrated fund manager” with a tweet snippet)

Original video