The Second-Order Rule: How I Find 300% Industrials

Let me tell you about the day I realized I was doing this wrong.

It was early 2025. I was staring at Nvidia for the hundredth time, thinking the same thing everyone else was thinking: the train already left. And it had. The obvious trade — buy the chips — was over before I ever saw it. I had no edge. Nobody in my feed had an edge. When your Uber driver is explaining the GPU supply chain to you, there’s no easy money left in the name everyone can name.

So I stopped looking at Nvidia. And I asked myself a dumb, almost childish question that has since become the only framework I use before I buy anything:

“Fine. Nvidia gets paid. But who is Nvidia — and everyone buying Nvidia — then forced to write a check to next?”

That question took me to FIX at $380. It took me to Powell at $40 (post-split). Neither one was a stroke of genius. They were two applications of the same rule, run mechanically, with no imagination required.

This edition doesn’t hand you a new ticker in the free section. It hands you something more useful: the rule itself. Once you see it, you can’t un-see it. And the free part is enough for you to start running it tonight.

The paid part is where the four operating filters live — the ones that turn the rule into a position you can actually hold — plus the three names that pass the test this week, with my entry levels. We’ll get there. Concept first.

Here’s the one thing the market genuinely does well: it prices the first invoice.

The first invoice is the name everyone knows. When the hyperscalers announce they’re going to spend north of $600 billion in 2026 — up roughly 36% year over year, with something like $450 billion of it aimed directly at AI infrastructure — the market knows exactly what to do. It buys Nvidia. It buys the chips. It prices the first invoice to perfection, often within hours.

The trouble is that “priced to perfection” and “good investment” are two different things. When a name is priced for perfection, you don’t get paid for being right. You only get punished if the world turns out to be less than perfect. And the world is almost never perfect.

So here’s rule zero, the one that comes before everything else:

Never buy the invoice the market has already read out loud.

This doesn’t mean Nvidia is a bad business. It’s an extraordinary business. It means your edge isn’t there. It’s further down the chain, where the story hasn’t been written yet.

Here’s the core of the framework. I’m going to walk it slowly, because the slowness is exactly where the edge lives.

Every large, forced expenditure — not discretionary, forced — sets off a cascade of downstream invoices. Most investors stop at the first one. The game is to walk down the cascade, invoice by invoice, until you hit a point of physical frictionthat nobody has turned into a story yet.

Let’s walk it together, using AI:

1. Invoice 1. Hyperscalers spend → they buy GPUs → Nvidia. Everyone knows it. Priced.

2. Invoice 2. Those GPUs draw an absurd amount of power → you need energy, cooling, electrical distribution → Vertiv, Eaton, the data-center REITs. The market built this story in 2024–2025. Largely priced now.

3. Invoice 3. That electrical distribution needs switchgear, transformers, medium-voltage breakers → Powell, nVent, GE Vernova. The market is only just starting to write this one. Some names have run; others haven’t.

4. Invoice 4. Those transformers need a very specific electrical steel (grain-oriented, “GOES”) that only a handful of mills on earth can produce — and European supply just got cut. Now we’re somewhere almost nobody is looking.

5. Invoice 5. That steel mill needs a forged part, a furnace, an input that nobody else knows how to make…

See the move? Every step down gets you closer to two things the market hates analyzing: a physical bottleneck and a boring monopoly.

That’s the entire mechanism behind FIX. In April 2025, everyone was staring at Invoice 1. I was looking at Invoice 3: who physically installs all the mechanical and electrical plumbing inside these data centers? The answer was a century-old contractor out of Houston that the market had just dumped because a Chinese startup released a cheaper model. Entry at $380. The thesis never had anything to do with the Chinese startup. It had to do with the fact that the backlog was filling faster than the company could hire.

Today the same rule keeps turning one notch lower. And that’s where most people make the mistake I want to help you avoid.

Here’s the honest part. I’m giving you the concept for free because it is worthless without discipline.

Most people will read the above, nod, feel smart for twenty minutes — then go buy the obvious name anyway. Or worse: they’ll walk too far down the cascade, find a name so obscure it’s illiquid, broken, or whose “monopoly” actually has a hidden substitute three countries away, and get trapped.

Walking down the cascade is easy. Knowing which notch to stop at — that’s the real work. And it’s exactly what separates a 300% industrial from a value trap with a great story.

FIX went from roughly $380 to over $1,700 as I write this. Powell, entered at $40 post-split, with a record backlog and a book-to-bill comfortably above 1, has done a multiple. These aren’t luck. They’re names that passed four filters — four questions I ask every candidate before I put a dollar in, and that kill about 90% of what the cascade surfaces.

Those four filters, plus the three names that pass them right now, are on the other side. Here’s exactly what you unlock:

The 4 operating filters of the Second-Order Rule. The exact grid I run. The physical-bottleneck filter (how to tell a real supply constraint from an imaginary one). The “not yet priced for perfection” filter (where I draw the line precisely — why I own Powell but passed on Vertiv at $339). The optionality filter (I won’t buy any name whose entire thesis dies if AI capex flatlines in 2027). And the hold filter: “can I hold this through a 30% drawdown without reopening the thesis?”— the one that’s saved me from my worst mistakes.

The scoring grid, to run on your own ideas. A reproducible table to score any candidate against all four filters. You don’t walk away with a fish. You walk away with the rod.

The 3 names that pass the test this week. Tickers, exchange, my entry levels, and the exact notch of the cascade each one sits on. All three are on Invoice 4 or lower — where the story isn’t written yet.

The 2 traps in the rule. The two setups where the cascade lies to you: the false bottleneck (the one with a hidden substitute) and the structurally declining monopoly that looks like a defensible one. I lost money on both. You don’t have to.

The honest bear case — on the framework itself. The scenario where this whole approach stops working. You should know it before you size a single position.

The annual subscription is currently $100. I’m raising it next month. If you’ve been reading the free editions waiting for the right one to go premium on — this is it. Because it isn’t a pick. It’s the method behind every pick.

Subscribe to Solo Capitalist Premium

None of the above is investment advice. I’m not your financial advisor. I write about what I do with my own money; you’re responsible for your own decisions and your own research. Past performance tells you nothing about the future. Prices quoted are approximate and current as of the time of writing.

If you take only one thing away from the free section, take this: the market is efficient on what it names, and lazy on what it doesn’t. Your edge isn’t being smarter than the market on Nvidia. It’s being more patient than the market, one notch lower, where the analysis is boring and the story doesn’t exist yet.

And there’s a structural reason this framework has years left in it, not months. An estimated 30–50% of the data-center capacity planned for 2026 is expected to slip into 2027–2028, stuck in power-grid interconnection queues. Translation: demand on the downstream invoices isn’t going to fall off next year. It’s going to stretch out. The names that worked in 2024–2025 worked because the buildout had begun. The names that work in 2026–2027 will be the ones that capture the continuation of that buildout — after the easy money has stopped moving, and the durable margin expansion has started.

That’s where I hunt. One notch below consensus. Always.