Actionable insights — My Investing Plan For The Next 5 Years

2:08 1. Classify every AI name as a seller-into-scarcity or a buyer/monetizer

1. For any AI-exposed company ask one question: is it selling the scarce input (a bottleneck), or is it a buyer that will monetize AI across its own products?
2. Map the seller hierarchy by how close it is to the bottleneck: accelerators → foundry → lithography → memory → networking → power/cooling → land/construction. Closer to the irreplaceable bottleneck = more durable pricing power.
3. Recognize the regime: in the scarcity phase the sellers carry the gains, so you can't sit out entirely — but the seller's edge is temporary by construction (it lasts only while demand > supply).

• Where a name sits on the bottleneck hierarchy; whether its earnings come from a transient shortage or a recurring relationship.

6:06 2. The cyclical-vs-structural durability test for hot sellers

1. For each seller running hot, ask what happens to its earnings when supply catches up to demand (phase 2 normalization) — the "wheat from tares" question.
2. Score durability on: one-time sale vs decades-long install + service contracts; recurring/annuity revenue; switching costs; ecosystem lock-in (e.g. CUDA).
3. Distrust the "it's no longer cyclical" story that always appears after a cyclical stock has multiplied — that belief is itself the warning sign.
4. Keep only durable sellers; treat the rest as rentals to avoid, not holdings.

• A cyclical stock at a record multiple with a "this time it's structural" narrative; the absence of recurring/annuity revenue under a hot seller.

11:42 3. Own SOME phase 1, but concentrate in the owners-of-the-customer

1. Hold a small foot in a durable phase-1 seller so you participate in the scarcity rally — don't be fully out.
2. Put the bulk of capital in phase-3 "buyers" that own the customer relationship and have many ways to monetize AI for 10–20 years (cloud, ads, search, messaging, productivity).
3. Use the depressed-metrics tell to your advantage: buyers look worse now because their capex is cash going out — that's the discount, since the same spend becomes long-lived profit later.

• Heavy-capex platforms with deep daily-use distribution; the rotation when seller pricing power fades (his ~1–2 year horizon).

16:15 4. Split de-rated software into rerate-up vs permanently-cheap

1. In a sector where multiples have already reset (30s → teens, stocks down 30–60%), don't buy the basket — split it.
2. Rerate-higher bucket: owns distribution, proprietary data, customer relationships, workflows, is a system of record — AI makes the product more useful and protects the moat.
3. Permanently-cheaper bucket: the product is "mostly just a UI feature / workflow shortcut" that AI agents can copy, bundle or bypass → investors assign a permanently lower multiple.
4. Apply the bundling test explicitly: can a giant give this feature away inside a suite the customer already pays for?

• A single-feature product a bigger platform can bundle for free; vs a system-of-record with proprietary data and switching costs.

18:29 5. Find the AI-insulated compounders the framework leaves alone

1. Separate "AI threat" from "AI noise" — many quality businesses are mostly non-AI and get dragged down with the sector for no structural reason.
2. Test for a real-world network/marketplace or habitual-use moat that AI doesn't erode (riders+drivers, hosts+guests, merchant tooling, daily-habit apps).
3. For the debatable ones, locate the true source of stickiness — if it's behavior/community rather than the content itself, the moat survives commoditized inputs.

• Marketplaces with two-sided network effects; apps whose stickiness is the habit/community, not the easily-copied content.

23:12 6. When the brand IS the business, judge brand-dilutive moves harshly

1. Identify companies whose entire premium rests on an intangible brand/heritage rather than measurable performance.
2. Weight management moves by their effect on that intangible, not just on the addressable market — a TAM-expanding product that cheapens the brand can be net-negative.
3. Watch the market's first reaction as a tell, then decide whether it under- or over-states the long-run brand damage.

• Luxury/heritage names chasing volume with off-brand products; a down-move that the market may be under-pricing on brand grounds.

17:17 7. Behind a headline blame story, find the structural mover

1. When a business failure is pinned on a loud, obvious cause, look past it for the slow structural force actually doing the damage.
2. Quantify the structural trend (revenue base, cost base, audience migration) to see if the decline is terminal regardless of the headline cause.
3. Trace who benefits from that structural shift — that's often the more durable, investable conclusion.

• "It's because of [person/politics]" narratives masking a secular platform shift; the platform that is quietly capturing the migrating audience.