1. Pick-and-shovel layer
These companies sell what everyone needs before AI apps can exist: GPUs, memory, networking, power, cooling, foundries, equipment and data centers.
Often strongest early-cycle
Lesson 1
First mental model: follow the AI dollar
2. Platform layer
Cloud providers, model labs, data platforms and security vendors turn raw compute into usable infrastructure for companies.
Durable, but valuation-sensitive
3. Application layer
Enterprise software, agents, automation, robotics and vertical AI tools capture value only if customers pay for productivity gains.
More stock-picking required
From your transcripts
Two useful lenses from Jordi Visser and Tom Nash
Jordi lens: scarcity and bottlenecks
AI is not just software: The transcript frames AI as a physical build-out: chips, data centers, power, cooling, chemicals, optical, memory, logistics.
Scarcity beats abundance: When demand jumps faster than supply, bottleneck suppliers can see huge earnings — but also parabolic moves and speed crashes.
Rolling bubbles: Leadership rotates. Last year’s winners can pause while new bottlenecks run. Avoid assuming one AI stock always leads.
Inflation/rates matter: If AI capex creates scarcity and inflation, higher yields can pressure valuations even when the AI story is real.
Tom lens: price down, business up
The best setup: Price down while the business is improving: revenue up, margins up, no key customer losses, no balance-sheet stress.
Ask why the stock is down: A sector panic can be opportunity; a broken business is usually not.
Buy slowly: Use a system, dollar-cost averaging, and pre-written rules instead of chasing headlines.
Misunderstanding creates alpha: Markets may mislabel companies: Palantir as consultancy, Salesforce/ServiceNow as AI victims, Snowflake as old database.
The AI vertical map
Where public-market AI opportunities sit
1Raw constraintsPower, grid, uranium/gas, cooling, land, permits, chemicals
2Semiconductor supply chainFoundries, lithography, equipment, packaging, memory, storage
3Compute & networkingGPUs, ASICs, Ethernet, optical, servers, data centers
4Cloud & model platformsHyperscalers, model APIs, AI development tools
5Data, security & governanceDatabases, observability, cyber, identity, compliance
6Applications & autonomyEnterprise software, agents, robots, autonomous vehicles, healthcare
Current market snapshot
Which AI verticals look crowded vs depressed?
This is a starting screen, not a buy/sell signal. “Overbought” means price action and valuation are hot. “Depressed” means the group is down or disliked enough to deserve fundamental work.
Shares dashboard
Representative AI shares by vertical
| Ticker | Company | Vertical | 1Y | 6M | From high | P/S | Revenue growth | Setup |
|---|
Your beginner process
A simple process for Edson: study first, buy slowly, avoid forced selling
- Learn the stack. Every stock belongs to a layer. Know which layer you are buying.
- Separate company quality from stock price. A great company can be a terrible buy if already priced for perfection.
- Use the 4-zone test. Best zone: price down + business up. Avoid price up + business down.
- Check seven fundamentals. Revenue growth, margins, free cash flow, balance sheet, customer retention, leadership stability, guidance.
- Control position size. AI can speed-crash. Consider watchlists, small starter positions, and written DCA rules.
- Respect your liquidity. Because your cash buffer is important, never let AI excitement create forced-selling risk.
Next lessons to add
Future content roadmap
Lesson 2
How to read earnings for AI companies.
Lesson 3
Valuation basics: P/E, P/S, free cash flow, TAM.
Lesson 4
Building a 10-stock AI watchlist.
Lesson 5
Dollar-cost averaging rules and risk limits.