Skip to content

P3-C1 · Real Hedge Fund AI Analysis Process

Core One-Liner

Analyzing AI isn't about Twitter — real hedge funds spend 6 months researching, hold for 24 months. The process is public, but 99% of retail investors don't follow it.

Real Analysis Process — Hedge Fund / Buffett / Finding Industry Bottlenecks / Multi-PM / Anti-thesis

P3-C1 (Part 3, Chapter 1 of 5). After this chapter, you can evaluate your own thesis using Coatue / Tiger Global's public 5-step process, breaking out of the retail investor decision-making model.

1. The Problem: Retail vs. Hedge Fund AI Analysis — Completely Different Processes

Dimension Retail Hedge Fund (Coatue / Tiger / Druckenmiller)
Information Source X / Twitter / KOLs / Headlines 13F + expert calls + IR + 8 quarters of earnings
Decision Cycle Within 1 week 3-6 months of research
Position Building One-time all-in 6-12 weeks of gradual build
Hold Period 1-3 months 18-36 months
Monitor Frequency Check stock price daily Quarterly review
Exit Trigger Noise / Panic Thesis invalidation / Extreme valuation

→ Same information, different density; same decision cycle, different results. Retail investors don't lose on information — they lose on process.

The 5 steps in this chapter are all public — Coatue discussed them at Bloomberg Invest 2024 / EMW 2024, Druckenmiller described them on Norges Bank "In Good Company" podcast (hosted by Nicolai Tangen) + Bloomberg Talks. 99% of retail investors know them but don't follow them, because of insufficient patience + insufficient time.

2. The Solution: Real Hedge Fund 5-Step Process

Step What It Does Time Can Retail Follow?
1. Universe Screening 8000+ companies → 300 watchlist Ongoing (monthly refresh) ✓ Yes (1 hr/month)
2. Deep Dive 30+ expert calls + earnings + supply chain 3-6 months / company △ Partially (use free transcripts instead of expert calls)
3. Position Build 6-12 weeks of gradual build 1-3 months ✓ Yes (change mindset)
4. Hold + Monitor Quarterly review, no daily price checks 6-36 months ✓ Yes (mental training)
5. Exit / Rebalance Thesis invalidation OR extreme valuation 1-12 months slow exit ✓ Yes

5 steps linked together: From looking at one stock to closing the position, average cycle is 1-3 years.

3. How It Works: 5-Step Breakdown + Real Public Cases

3.1 Step 1 — Universe Screening

Coatue mentioned at Bloomberg Invest 2024 / EMW 2024 that they maintain a ~300 company watchlist. Out of 8000+ public US stocks, they only focus on:

  • Market cap > $5B (liquidity + institutional investability)
  • TMT sector (Tech/Media/Telecom — Coatue's 25-year specialty)
  • Understandable business (no biotech / commodity / complex finance)
  • Financial quality (gross margin > 50% or strong unit economics)

Monthly refresh: New IPOs / earnings surprises / macro shifts get added or removed.

Public case: Coatue added CRWV (CoreWeave) to their watchlist in 2023 — well before its IPO. Because they track the entire AI infrastructure chain, they were ready when neocloud emerged.

Retail comparison: Retail investors have no watchlist. They buy when they see a recommendation on X. Coatue's watchlist is the prerequisite for the next 6 months of research.

3.2 Step 2 — Deep Dive (3-6 months / company)

Out of the 300 watchlist, only 5-10 are selected for deep research in any given quarter. For each company:

Research Action Retail Alternative (Free)
30+ expert calls (ex-employees / customers / suppliers, via Tegus / GLG / Guidepoint) Find ex-employees on LinkedIn / use Glassdoor / Blind for chatter
5+ supply chain checks Read NVDA earnings comments on SK Hynix, read ASML earnings on TSM
8 quarters of earnings + transcripts Motley Fool free transcripts (NVDA / TSM / MSFT all available)
Major trend cross-comparison Look at 13F cross-section (10 top funds all adding = strong signal)

Public case: Before Druckenmiller added NVDA in 2023, he described a "weeks of expert calls before buying" research process on Norges Bank "In Good Company" / Bloomberg interviews (paraphrase, not verbatim). He didn't buy just because ChatGPT was hot.

Retail comparison: Retail investors don't do expert calls; they use Twitter as their expert. Information density is 100x different. But actually reading free transcripts + 13F + IR PDFs already puts you ahead of 90% of retail investors.

3.3 Step 3 — Position Build (1-3 months of gradual build)

Research done, decision to build a position. Not all-in at once, because:

  • Large funds buying $1B at once pushes the price up — they'd be buying their own inflated entry
  • Gradual building allows cost averaging (average entry price)
  • Leaves room to change your mind — new data mid-process might invalidate the thesis

Coatue's public pattern: starter 1% → ramp 2-3% → core 5-8%. Full position in a single stock < 10% (concentration limit).

Druckenmiller NVDA Replay (counter-example: even pros sell too early)

Date Action Rationale at the time
2022-2023 Build NVDA + MSFT "AI is real, looking out 4-5 years"
2024 Q1-Q3 Fully exited "Valuation too high, need to take a breather"
2024/10 Bloomberg Publicly admitted "big mistake" "Sold too early; you only see this kind of company a few times in a lifetime"
2025/1 post-DeepSeek — (no position)

→ The lesson is not "hold to zero", but: (1) don't sell on valuation alone if the thesis hasn't broken; (2) on big-trend opportunities, missing the upside costs more than a -30% drawdown; (3) even a legend like Druck makes mistakes — your discipline matters more than your forecast accuracy.

Retail comparison: Retail investors go all-in at once. No room to trim. One mistake and it's a heavy loss. Starter → core over 3-4 weeks — a mindset shift you can learn.

3.4 Step 4 — Hold + Monitor (6-36 months)

After building the position, enter monitor mode. Don't check the stock price daily. Instead:

Review Dimension Frequency
Earnings vs thesis (supports still valid? red_flags triggered?) Quarterly (4 times/year)
Industry shifts (hyperscaler capex / customer quality / regulation) Monthly
Position size adjustment (price changes cause drift, rebalance) Quarterly
Price (intraday) Don't check

Coatue public: Average hold 18-24 months. No short-term trading.

Coatue NVDA Replay (actual: 2024 large trim)

Date Action
2023 Built NVDA position
2024 Q1-Q4 Trimmed 77-80% (13F public data)
2025 Partially rotated into other AI infra (ANET / AVGO etc.)

→ Coatue's actual pattern is rotate (not hold to infinity): once NVDA's thesis is priced in, rotate to the next-wave beneficiary. This is also rational fund behavior, just different from Druck's full liquidation.

Retail comparison: Retail investors check the stock price daily. Down 5% and they sell. Up 30% and they're anchored, won't sell. Psychology driven by noise.

3.5 Step 5 — Exit / Rebalance (1-12 months slow exit)

Three exit triggers:

  1. Thesis invalidation — supports no longer hold, red_flag triggered
  2. Extreme valuation — fwd PE above 5-year 90th percentile + growth slowing
  3. Better opportunity — capital reallocated to a new thesis

Coatue public: Exits are also gradual, over weeks to months. Not a one-time liquidation.

Public case: Coatue reduced META in 2023 (Reality Labs burning cash + soft ad market). Re-entered in 2024 (RL Labs cuts + AI acceleration). This slow exit + re-entry is institutional norm.

Retail comparison: Retail investors liquidate and never look back. Coatue tracks for 5 years, re-entering with precise timing.

4. vs. Retail Approach — 5-Dimension Full Comparison

Dimension Retail Hedge Fund What You Can Change
Information Source X / KOLs 13F / Earnings / Transcripts ✓ Use free Motley Fool / SEC EDGAR / Whalewisdom
Decision Cycle 1 week 3-6 months ✓ Force yourself to read at least 4 quarters of transcripts before deciding
Position Building One-time all-in Gradual over 1-3 months ✓ Starter (30% target position) → ramp → core
Hold Period 1-3 months 18-36 months △ Mental training — at least 6 months before reviewing
Monitor Frequency Check price daily Quarterly review △ Tool changes (delete app / turn off price alerts)
Exit Noise-driven Thesis-driven slow exit ✓ Write invalidation_triggers in your thesis

Core insight: Not a single step in the 5-step process is a secret. All top PMs discuss it in public speeches. Retail investors don't follow it not because they don't know, but because of gaps in patience / time / psychology.

5. Try It: Evaluate One of Your Stocks Using the 5 Steps

Task (~2 hours, but can be split over 5 days, one step per day):

Step What You Can Do
Step 1 Does your ticker meet Coatue's watchlist criteria? (Market cap > $5B / TMT / gross margin > 50%)
Step 2 Have you read 8 quarters of Motley Fool transcripts? Found 1 ex-employee / customer on LinkedIn?
Step 3 Is your position in starter / ramp / core phase? Does it match your target position size?
Step 4 When was the last time you checked this stock's price? (Ideal answer: more than 1 week ago)
Step 5 Does your thesis have explicit invalidation_triggers?

Self-check (3 items checked → proceed to P3-C2):

  • You can find at least 1 source deeper than Twitter (e.g., Motley Fool full transcript)
  • You can extend your holding decision cycle from 1 week to 1 month (force yourself to read 4 quarters)
  • You can build a position starter → core gradually (no more one-time all-in)

6. What's Next

You've seen the hedge fund process — they hold moderately concentrated portfolios (10-30 positions). Buffett is more concentrated — 5-10 holdings, hold for 10+ years. Different framework.

→ P3-C2 · Real Buffett-Style AI Analysis transplants Buffett's 5-step moat analysis (using AAPL as a case study) to AI stocks.

7. Deep Dive (optional): 5 Free Public Sources to Learn from Top PMs

Click to open public source map

Stanley Druckenmiller (Duquesne Family Office, AUM undisclosed but estimated ~$3-4B per public 13F): - Norges Bank "In Good Company" podcast (hosted by Nicolai Tangen, 2024-11) — deep dive on macro + AI - Bloomberg Talks 2024-10 — Fed / NVDA / election - How Leaders Lead w/ David Novak — process + career → Learn his macro top-down + concentrated bets + ride winners (but trim at extreme valuation)

Coatue Philippe Laffont (~$30B public 13F / $58B total AUM per Form ADV): - Bloomberg Invest 2024 (public speech w/ Rubenstein, "$100T will be invested in AI") - Coatue EMW 2024 / 2025 (annual East Meets West "State of Markets" public deck) - CNBC Delivering Alpha 2024 → Learn their AI mosaic framework (simultaneous allocation across upstream, midstream, downstream)

Tiger Global Chase Coleman ($25B+ AUM): - 13F holdings public (Whalewisdom / Dataroma free) - Chase Coleman himself rarely public, follow via 13F → Learn Tiger Cubs cross-section (Lone Pine / Maverick / Whale Rock are all Tiger lineage)

Howard Marks (Oaktree) ($170B AUM): - Public memos (oaktreecapital.com free, 1-2 per month) - Not AI-specific but monthly market cycle commentary trains mental models → Learn his cycle awareness + risk-first thinking

Bill Ackman (Pershing Square) ($15B AUM): - Quarterly letters (pershingsquareholdings.com free) - Active on Twitter (but filter noise) → Learn his concentrated activist + public thesis writing style

Public source priority: 1. Company IR PDF + SEC 10-K/10-Q (most ground truth, completely free) 2. Motley Fool transcripts (earnings transcripts available free 1-2 days after) 3. 13F (Whalewisdom / Dataroma) (institutional holdings, 45-day lag) 4. Public podcasts (Lex / Acquired / All-In / Invest Like the Best) 5. PM public letters (Buffett / Marks / Ackman / Druckenmiller occasional interviews)

These 5 sources are completely free and contain 90% of the information a retail investor needs. No need for any paid substack / SemiAnalysis / Stratechery — those are incremental edges, not the base.

8. Further Reading (this chapter — hedge fund 5-step process)

All free sources, aligned with P5 0-paid policy

Books (borrow from library):

  • Sebastian Mallaby "More Money Than God" (2010) — 60 years of hedge fund history + Soros / Druckenmiller / Robertson
  • Maneet Ahuja "The Alpha Masters" (2012) — process interviews with 9 top PMs (Ackman / Tepper / Einhorn etc.)
  • Steven Drobny "Inside the House of Money" (2006) — process portrait of global macro PMs
  • Anthony Scaramucci "The Little Book of Hedge Funds" (2012) — industry basics + strategy taxonomy

Classic papers / primary sources:

Wikipedia (3-10 min):

Videos / lectures (top PM primary interviews):

Podcasts (1-3 hr/episode):

Company letters & memos (primary, free):

Pair with chapter self-check: see the 5-step ticker evaluation in section 5 above.