The Focus List: Own the Receipts, Not the Frontier

This is our first ever Focus List. The idea is simple: stop chasing the AI frontier. Own the companies getting paid for it right now.

If you opened your brokerage app on Monday, you saw the split screen everyone on Wall Street is talking about.

On one side: the Fed. After three years of holding, they hiked. Unanimous. 12–0. A quarter point to 3.75%–4%, with one more penciled in before year-end. The 10-year Treasury closed just over 5%. That’s not an abstract number. That’s your savings account, your car loan, your credit card bill.

On the other side: the AI debate finally got loud. Over the weekend, Anthropic’s CEO Dario Amodei published an essay basically saying, “maybe we should all slow down building frontier AI.” Two rival labs publicly agreed.

The market picked a side immediately. The semiconductor ETF fell 5.6% in one day. Broadcom fell 4.8%. Constellation Energy — which powers the whole thing — fell 7.1%. And Netflix? Netflix rose almost 4%.

One side sells the shovels for AI. The other sells an escape from it.

That’s what this list is about. We’re not betting on who wins the frontier. We’re buying the four companies that are already collecting rent, with real earnings, that are suddenly trading like they’re on sale.

Every one of them is doing more business than ever. Every one of them is trading at a fraction of its old multiple. Broadcom at roughly 17x next year’s earnings versus a five-year average of 32x. Netflix at roughly 20x versus an average near 40x. Constellation a third below its high. SoFi down more than a third since January.

This is Edition #1, so we’re not scoring anything yet. Going forward each name will have four things: why it’s here, what it has to prove, what breaks the thesis, and what would make us flip.

Let’s get into it.

1. Broadcom (AVGO) — The Arms Dealer

Think of Nvidia as selling off-the-rack suits. Amazing suits, everyone wants one. Broadcom is the bespoke tailor.

Google, Meta, Apple, ByteDance, OpenAI and Anthropic don’t all want the same chip. They want their own chip, designed for their own data centers. Broadcom designs it for them. Then it also sells the switches and fiber optics that connect those chips together. It owns roughly 70% of the AI ethernet switching market. It’s an arms dealer with six customers who can’t stop ordering.

And they are ordering. Last quarter Broadcom did $29.6 billion in revenue, a record, up 86%. AI chip revenue was $16.7 billion, up 221%. Operating margin hit a record 68%. That’s not a software company margin, that’s a monopoly margin.

So why did it drop? Because Wall Street doesn’t care about last quarter. It cares about next quarter. Guidance came in about $300 million light. The stock fell 3% after hours even though management told you they see $58 billion in AI revenue this fiscal year, roughly $115 billion next year, and a line of sight to $230 billion the year after that.

CEO Hock Tan was asked about the whole “slow down AI” debate. His answer was basically: nothing has changed in my order book.

Here’s the catch, and it’s a real one: those are targets, not cash in the bank. And Broadcom is so confident it set up a financing platform with Apollo and Blackstone to help OpenAI and Anthropic pay for more than 20 gigawatts of compute. Translation: “If you can’t afford the chips right now, we’ll help finance it.” It’s bold. It also means if two cash-burning customers hit a wall, Broadcom feels it.

What we’re watching: the December print. We want to hear that $115 billion number for 2027 repeated, and that margin held. If they start hedging, the story changes.

Our take: Accumulate. The compression already happened. You’re getting the arms dealer at half its old multiple.

2. Constellation Energy (CEG) — The Landlord

If Broadcom is the arms dealer, Constellation is the landlord who owns the only building with power.

This is the largest nuclear fleet in the US, part of a 55-gigawatt generating fleet of always-on, already-licensed power. You can’t just build that. It takes a decade, billions of dollars, and a miracle from regulators. AI data centers need that power yesterday, especially in places like Virginia and Illinois where the grid is already tapped out.

That’s why Meta signed a 20-year deal for 1,100+ megawatts from Constellation’s Clinton plant. Why Microsoft signed to bring Three Mile Island Unit 1 back online. These are 20-year contracts. That’s as close to guaranteed rent as you get in energy.

But here’s why the stock got smoked — down 7.1% on Monday, sitting in the bottom sixth of its 52-week range. Two things new investors need to understand.

First, rising rates punish stocks like this. Big, capital-heavy, long-term projects get less attractive when the 10-year is at 5%.

Second, and more important, none of that big tech money shows up this year. The Meta deal starts June 2027. The Microsoft deal starts when Three Mile Island restarts in 2027. Some deals start in 2029. So you’re paying roughly 20x earnings today for growth that is contractually scheduled to slip if regulators move slow.

There’s also a weird grid fight happening right now. PJM, the grid operator for 13 states and Washington, DC, just admitted it came up roughly 6,800 megawatts short. It wants to pay new plants a premium to build fast — up to $555 per megawatt-day, against the $325 cap that binds Constellation’s existing fleet. Only new generation can bid for that premium. Constellation’s existing plants are not eligible.

What we’re watching: we don’t need a perfect earnings print in November. We need two headlines: the Crane restart — Three Mile Island Unit 1 under its new name — confirmed for 2027, and the December capacity auction. If that auction clears at the cap again with almost no new supply, it proves how desperate the grid is, and how valuable Constellation’s fleet is.

Our take: Wait for a better entry. We love the asset. It literally cannot be replicated. But we’d rather buy it when the market remembers that the rent starts in 2027, not 2026.

3. SoFi (SOFI) — The Bank in Your Pocket

SoFi is a fully licensed bank that just happens to live in your phone. 15.8 million members. Personal loans, student loans, home loans, investing, crypto. And it also owns Galileo, the tech platform that a bunch of other fintechs build their apps on top of.

It’s on this list because of the Fed. SoFi has to pay you interest to keep your deposits. When the Fed hikes, it has to pay you more, immediately. But the loans it already made are fixed at the old, lower rate. So its profit margin gets squeezed from both sides.

The business itself is humming. $1.22 billion in revenue last quarter, up 43%. 11th profitable quarter in a row. Record loan originations. Earlier this month it got regulatory approval for its own stablecoin issuance, and Kraken joined its settlement network.

But the stock is the most beat-up of the four: down more than a third since January, 48% below its high, trading below its 200-day average. Because the story Wall Street wants — that SoFi is a high-margin tech platform — isn’t what the numbers show yet. Lending was $724 million, 59% of revenue, growing 63% at a 56% margin. The Tech Platform, the part that’s supposed to justify a software multiple? It shrank 23% to $84 million at a 14% margin.

The smallest, slowest, least profitable part is being asked to carry the whole valuation. At 28x this year’s guided earnings, that’s a lot to ask for a consumer lender heading into a hiking cycle.

What we’re watching: two numbers in late October when it reports, right when the Fed meets again. Net interest margin back above 6%, and Galileo growing again year-over-year. Also, does fee-based revenue finally become more than half the business? Right now it’s 39% of revenue.

Our take: Wait. The rate cycle just turned against a deposit-funded lender. We want to see margin recover before we get excited.

4. Netflix (NFLX) — The Escape Hatch

Netflix is the one name on this list that goes up if AI spending goes down.

It doesn’t buy custom silicon. It doesn’t need nuclear plants. It buys scripts. It sells subscriptions and ads in 190 countries.

That’s why it was green on a day semis were down 5.6%. It’s the ultimate rotation name. If investors decide they don’t want to pay for the frontier anymore, they park their money somewhere that benefits from people staying home and streaming.

And it’s stupid cheap historically. About 20x forward earnings versus a five-year average near 40x. That’s the biggest de-rating of the four, on record profit and record free cash flow guidance of $12.5 billion.

The last print had a little crack though. Revenue of $12.56 billion slightly missed and the stock fell 7% the next day. Viewing hours in the first half were up only 2%, with hours per member a bit softer. They already raised prices across all plans earlier this year. So how do you grow 12% when people are watching the same amount? You make ads work. They’re tracking to $3 billion in ad revenue, but now they’re competing with Amazon, Google, and a combined Paramount–Warner Bros. library.

What we’re watching: October 20. That’s the only company-confirmed date on this list. We need revenue at or above $12.86 billion and operating margin at 33.2%.

Our take: Accumulate. It’s not an AI winner. That’s the point. It’s where you hide while everyone else argues about capex.

The bottom line

The market spent last week asking, “Do we still want to pay for the frontier?”

We’re answering a different question: who’s already getting paid for the receipts?

Broadcom shipped the compute. Constellation contracted the power. SoFi originated the loans. Netflix collected the subscription.

That’s the Focus List for this week. We’ll track these four and tell you when our call changes.

Analysis, not advice. We may own these names and we may sell them at any time.