Most AI investing conversation is about who's winning. I find that less interesting than the question of what displacement actually does to everything else. The infrastructure it requires, the business models it breaks, the sectors it forces to adapt, and the few names where the market is still late.
That's the core of what I call the AI displacement thesis. Not a prediction about which model dominates or which hyperscaler owns the stack five years from now. A framework for finding what I'd describe as mispriced future importance. Names where the upside can be large, the path to that upside is at least partly visible, and the evidence is stronger than the narrative.
What I'm actually trying to find
The watchlist isn't a buy list. It's a decision-support tool built around a specific type of opportunity: grounded speculation.
Speculative today, but with an identifiable mechanism for becoming real.
That distinction matters. Pure speculation is a narrative with no audit trail. Grounded speculation has concrete evidence that the thesis is moving such as contracts, procurement, partnerships, hiring patterns, insider behaviour, operating signals. The speculation is in how big it gets and how fast. The grounding is in whether it's actually happening.
This is why I don't start from market cap and work down. The highest market cap names in AI are the most crowded trades. The rerating on Nvidia at current valuations requires a level of continued dominance that deserves real scrutiny. I'm more interested in where the market is still late.
The signal hierarchy
When I'm assessing a name, I weight evidence in a specific order.
- Insider alignment: open-market buying by founders or executives, insiders holding through volatility rather than selling, high long-term founder ownership. This matters more than almost anything else because it's one of the few signals that's genuinely hard to fake at scale.
- External validation: official contracts, partnerships, government procurement, formal distribution deals. These are the mechanisms that convert a thesis into a visible catalyst path.
- Operating breadcrumbs: unusual capex moves, hiring patterns that don't match the stated strategy, segment reporting changes that suggest something is being repositioned.
The framework strongly prefers pre-obvious validation over pure momentum. By the time a thesis is consensus, the easy rerating is probably done.
The seven lenses
The watchlist organises names across seven themes, and it's worth being clear about what each one is actually tracking.
- Power and cooling is the most structural of the lenses. The compute running AI models is extraordinarily power-hungry, and the grid investment required to support that is a decade-long cycle. This isn't a speculative bet on which model wins. It's a near-certain consequence of the direction of travel, and the utilities, cooling specialists, and grid technology companies sitting in that path benefit regardless of the AI race outcome. I'm particularly attentive to where renewable build-out intersects with data centre demand in ways that global commentary mostly ignores.
- AI infrastructure is the picks-and-shovels layer: semiconductors, networking, storage, cloud plumbing. The most crowded theme and the one that requires the most valuation discipline. Being essential to AI doesn't automatically make the current price a good entry.
- Deep tech covers companies doing genuinely novel things at the model and application layer that haven't yet hit mainstream attention. Higher risk, more potential asymmetry, held more lightly.
- Adapt is the one I find most compelling right now. Established businesses in disrupted categories making credible structural pivots. Not AI bolted onto a press release, actually restructuring the model around what AI makes possible. These are hard to identify and easy to get wrong, but the rerating when you find a management team that genuinely gets it can be significant. Management behaviour is a key signal here.
- Pain trade tracks the displacement candidates. Sectors facing structural pressure that the market may not have fully priced. Displacement is slower and messier than optimists expect, and incumbents fight longer. But I want visibility on where the pressure is building, and this sleeve also supports the short side of the book.
- Policy response is underrated as an investment lens. AI regulation is arriving across multiple jurisdictions and the companies best positioned to navigate compliance complexity, or that benefit from it as a moat, are worth tracking. Some defence and government tech plays sit here too.
- Escapism is the contrarian position. The thesis is that as AI colonises more of knowledge work and digital life, demand grows for things that are explicitly not AI: physical experiences, handmade and artisanal goods, analogue media, presence-based services. I'm watching for this to show up in revenue, not just sentiment.
How names get placed
Three buckets, with different evidence thresholds.
- Starter Now requires the strongest case: two signals plus a live catalyst inside three to six months, or three signals where insider alignment is one of them. These are the names I'm comfortable establishing a real position in.
- Monitor needs substance but with looser catalyst timing. Two signals, watching for the next proof point that would move it to Starter Now.
- Reach is higher-risk optionality: one strong signal plus a plausible catalyst path. Smaller positions, longer patience required.
The watchlist aims for around 20 to 30 names in total, with six to eight Reach positions allowed. It's intentionally concentrated in the strongest themes rather than balanced for its own sake.
What the framework dislikes
Pure hype with no evidence trail. Crowded positions where the next rerating leg looks fully priced. Names that are operationally messy without improving proof. Illiquid positions I can't realistically enter or exit. Companies where management promotes aggressively but insider behaviour doesn't match the words.
There's also a thesis-conflict penalty I apply. If I like a company but part of its rerating now depends on a sector dynamic I'm bearish on elsewhere, ad-tech exposure in a business I'm otherwise interested in, for instance, that creates internal conflict in the book that needs to be acknowledged and scored against.
What I'm not claiming
I'm a marketer and strategist, not a fund manager. This framework was built to help me think more rigorously across the impact of AI which could be seen through stocks.
I'm not claiming this produces alpha reliably. I'm claiming the framing, mispriced future importance, grounded speculation, pre-obvious validation, surfaces more interesting and less crowded questions than starting from market cap or chasing whatever the market decided to be excited about this quarter.
The displacement is happening.