More and more returns are being driven

by the themes a company is exposed to.

That's why thematic equity research is

so valuable. Today I'm speaking with a

clear leader in thematic research,

Catrini. Many of you are familiar with

Catrini's work. But in addition to

research, the Catrini team also has a

tool called the Catrindex to track

custom indexes and baskets that they

build. This helps investors track

performance, think through trade

expression, and improve portfolio

construction. Throughout this interview,

we're going to put up some charts from

the Satrindex so we can show you various

baskets that we're talking about. The

Satrindex subscription is a separate

product from Satrini Research and up

until recently, if you wanted to

subscribe to both, you needed to buy

them separately. But now, you can get

Satrini Research and the Satrindex

together in one bundle through Substack.

Check out my link in the description or

go to catrinsearch.com/mjack

for an exclusive 25% discount on this

bundle. The deal expires on January

14th. Let's get into it. joined once

again by an investor and analyst who I

respect a lot. I really like the way he

thinks about markets and I know many in

my audience do as well. Satrini, welcome

back to Monetary Matters. How are you

thinking about markets in 2026? For

subscribers, you recently wrote 26

trades for 2026. You have an incredibly

large and voluminous number of trade

ideas, but where are you thinking you're

going to be seeing the opportunity in

2026? And what is the difference between

something that's just on the bookshelf

versus something that is actually going

to be implemented in your model

portfolio for clients as well as you

personally?

>> Thanks for having me back, Jack. Uh, it

it every year that I do this, it

surprises me by the middle of the year

which trades I end up putting on, which

I don't. that normally what happens is I

have a couple that I'm really like this

is going to this is definitely going to

make it into the portfolio and then I

get halfway through the year and I look

at what actually makes it into the

portfolio and it's not that uh last year

we had a piece I think it was like the

16th trait that was like if you remember

you remember when there were all those

drones over in New Jersey

>> and we just kind of memory hold it. So

that was happening around the time we

were writing 25 trades. And I looked at

that and I said people are going to

start thinking about like

counter unmanned aerial surveillance,

how to defend against electronic warfare

and this stuff. And then around the Iran

escalation put on the quote unquote

drone basket, but it had nothing to do

with the drones in New Jersey. So it's

always interesting to see how the year

progresses, but there are a few that I

have pretty high conviction on so far.

But again, it's mostly a thematic watch

list. The reason I like doing it is just

because it forces you to look at things

that you otherwise wouldn't really think

about.

>> And so we when we did that interview

last year, 25 trades for 2025, we talked

a lot about themes. We talked about

electronic warfare or drones. We talked

about Ukraine Russia normalization. And

those themes were up 76% and 75%

year-to- date. and you actually you

reviewed all of the themes and a little

more than half of them outperformed the

S&P and more than 80% had positive

returns. So definitely helped by the

fact that we had a bull market but very

solid performance there. And I should

say also

>> point out that there were some dumb ones

in there too, right? Like just for

anyone that reads this and is, yeah, I'm

going to put on every single one of

these trades. Like the one of them was

we looked at like these remittance

companies and they were so cheap and so

enticing. I don't even remember what we

said, but I looked back at it when we

went to do the scorecard and I was like,

why did we think that remittances would

do well when Trump's doing like mass

deportations? That doesn't make any

sense. So again, it's something where it

the really the purpose of this is the

in our vocation, we spend so much time

thinking about being right about the

future. And you inevitably will run into

a point where you look at like this

happened for me with gold miners this

year where it was like I remember

looking at gold miners and they're so

cheap relative to where gold's at and I

just didn't do anything with it and it

was because I got so in the weeds on it

and I was like ah you know like a mine's

just a pole with a liar standing on top

of you. and I ended up deciding against

it. And I look back and I say, man, if I

had just done the obvious thing, which

is buy the gold miners because gold's

already up 40% of your day, that would

have been a great trade. You always find

yourself missing some obvious things and

that's inevitable. But for me at least,

the way that you can most effectively

defend against that is once in a while

you just allow yourself the time to not

feel the pressure of everything I do

needs to be right. I need to be skating

toward the and you just say what do I

think is going to happen? What what like

what are the kind of obvious? What are

the not so obvious? What's the biggest

list I can make of things that probably

could happen? And that's what this is.

It's like the one time a year where

we're not just neurotic about being

right.

>> Right. And um

you you also have a model portfolio, the

Citrindex that you track very rigorously

and that the Citrx I see for 2025 as we

record is up 22% year-to- date versus

18.5%

for the S&P 500. And since inception in

2023, this trend index is up 217% versus

69% for the S&P. that is a way to

actually track, you know, what's doing

because, you know, I could have my

investment newsletter and I throw out a

thousand ideas and 10 of them are going

to make me look like an absolute genius.

But actually determining the sizing, the

entry points, the exit points,

rotations, when to hold on to a winner,

like that is really what makes a great

investor, not just the ideas. So I think

that's important and monetary matters

listeners can get that. We're you're

selling that as a bundle. So the Catrini

research substack as well as the

Catrindex tracking tool where you can

track all of the baskets and that in

real time for that bundle. Monetary

matters listeners can get a 25% discount

until the middle of January. And I I

think James there there are some hedge

funds that are not tracking like with as

much rigor their positions as this is

trend deck. So even though it is a model

portfolio, I've been using it and I'm

definitely impressed impressed by it.

Tell tell us a little bit about that.

>> Well, like you said, you nailed it. The

e when you're writing research that the

easiest thing is is you know, you write

the research, the things that are wrong,

you never really talk about them again.

You know, you say like so and so stock

is a long and then it goes up and you

say, "Yeah, look, look how sick we did."

It's very rare, I think, and it's like

the much more difficult thing to do.

Obviously if you're managing a portfolio

eventually you take a draw down or you

underperform that happens right and when

anyone can log into a website and see it

it's you know like oh you're talking

about this but you you know you took a

draw down yeah well that happens but

it's for me the most effective way to

communicate whether I still think that

something's a good opportunity whether

whether we can always write about

something and then it's a different

matter entirely of is this just

interesting is this something that

you're actively buying is this something

that what's your time frame on this.

It's just a we're all reading research

because we want to make money in the

stock market. So, it's the mo for me at

least it's the most effective way of

communicating that. The other cool thing

that that I have focused on with Catrini

is creating these kinds of thematic

factors. If you think about um

artificial intelligence for example,

it's not necessarily the same as growth

or t. And there will be periods of time

where things find themselves into, you

know, quoteunquote AI factor that

wouldn't necessarily otherwise be

included. Obviously, you're going to

have all the data center stuff in there

uh in terms of semiconductors, but what

about when power becomes, you know,

something that everyone's talking about

and you know, Genova and Seammen's

energy uh find their way into that

factor and start trading with beta to

this AI factor. So, it has every single

basket that we've ever created and you

can and real-time tracking and you can

get a good feel on at least for our

thematic universe which now after 3

years is pretty deep like we have a deep

bench of themes. It's like a it's a very

interesting addition to just like

tracking your classic low ball growth

dividend whatever factors.

>> Seeing now 132 baskets that that is a

lot of basket. So, James, so talk about

the AI trade broadening. I'm reading

from your 26 trades piece. You write

that the being the surefire path to

being early to some of the most

profitable parts of the AI trade has

been looking at areas that currently

have little to no AI premium baked in

and reasoning out 6 to 12 months as to

whether they'll become a crucial part of

the supply chain. So basically buying

stocks that no one really associates

with AI but make products or services

that are related to AI that are going to

be in in high demand as the AI buildout

continues. So the topic that's been on

everyone's

mind as as it relates to AI is, you

know, where is the return on investment

coming from? And that's a the whole bag

of worms for the hyperscalers and the

companies that are selling into the data

centers and all that stuff. The

interesting thing that happened this

year, there's been a lot of focus on the

science fiction version of AI. When's it

going to cure cancer? When's it going

to, you know, um kind of reach AGI or or

whatever you want to call it. Now I

think right that that what happened this

year essentially is we found a

capability gap where the trade is going

to broaden out to not about what AI

might do in 5 years but very much about

what it can do already that companies

have been lagging and figuring out. So

the reality is essentially that that if

you think of any organization, there's

always going to be a subset of the kind

of organizational pyramid at the bottom

where it's pretty much undifferiated

labor and AI progressed in 2025 to the

level where the technology to replace

that already exists. Not necessarily the

people above, let's say, the bottom 20%,

but any organization has a certain

amount of employees that are creating

negative value. And uh there's a huge

gap between what AI can do today and

what most organizations are actually

using it for. And that's where I think

this new part of the trade lives. And

the reason why it's so interesting to me

is it's kind of similar to the robotics

thesis that we had which we spoke about

um where the automotive cycle was so bad

that you could buy these secular winners

from robotics at this kind of cyclical

trough price. I can give you a real

example like I was talking to a senior

person at a unnamed large professional

services firm recently and they were

telling me yeah you know we're

experimenting with AI we're letting

junior analysts use chatpt to summarize

PDFs we're using quen and all this stuff

but at the same time think about the

standard junior analyst at a bank or

consultancy what's their actual job it's

not strategizing it's taking a logo from

a PDF removing the background aligning

it perfectly on a PowerPoint slide at 2

in the morning so that their MD doesn't

yell at them that is what David Greyber

in his book called a quote unquote

job. He has this great quote

about John Maynard Kees saying in 1930

by the end of the century we'll all have

a 15- hour work week because technology

will have taken us so far that you don't

need to be in the office 40 hours a

week. And he was right. Technology did

progress to that level but we just

created a bunch more work that that

nobody really needs to do. And there's a

lot of people that show up to work like

email jobs and they don't feel that what

they're doing needs to be done. So right

now, like at this exact period of time,

an AI agent can do that same logo

aligning task for in 4 seconds for a

fraction of a penny. But Fortune 500

companies are still paying a guy from

Warden $150,000 a year to do it manually

because their own like internal

bureaucracy hasn't caught up. Technology

advances at this exponential curve and

human adoption of technologies is

relatively linear. So I think that

there's a trade there for sure. You want

to own the companies that have very

bloated organizations and show some

intent to cut that down with artificial

intelligence. And the interesting part

right now that's unique about this exact

moment in time, nobody's stopping them

from doing that. If you think about the

social implications of like mass

unemployment or that's actually

occurring, there's no eventually that

will cause an issue. But right now

there's no regulatory blocker, there's

no technical blocker. It's just that

these organizations don't know how to

rewire themselves. So you want when you

look at some of these companies like

that have been

trading like AI losers and that's gotten

their valuations extremely attractive.

If you take the idea of this company

might end up being an AI loser. Yes, AI

can code now. So maybe you don't need

Axenture, maybe don't need Cap Gemini,

but that's not going to happen

overnight. and the ad agencies, Omnicom,

WPP, they're not going to get replaced

by AI tomorrow, but they can replace a

very significant amount of their labor

pool with AI today. So, looking out

across this universe of stocks that

could really see a benefit from

organizations realizing that they can

finally utilize AI, that it's good

enough. When people talk about this

adoption being slow, I think that's the

wrong framing. The tech adoption is

fast, but the organizational adoption is

glacial. And

there's an interesting universe of

stocks that are just very cheap that

that are traded like AI losers and could

realistically cut half of their

workforce and SGNA would would go down.

Their margins would go up pretty

significantly relative to their peers

and then they'd rerate and I think

that'll happen in 2026. If you think of

the other interesting angle here is you

remember we we spent a lot of time in

the 2010s talking about the cloud

transition

and it was a big deal that companies had

all this data that was sitting in analog

whether it's paper or whatever notes and

that cloud transition we don't talk

about it that much anymore but it never

fully happened. So the second that one

of these companies manages to utilize AI

to cut a significant portion of jobs and

ends up being fine if not better more

lean realizing better margins their

competitors are going to say oh we want

to do that too and the bottleneck is

going to become we don't have that much

we're not at the level where we can be

able to do that we like the companies

that assist with this think about like

SAP like they haven't gotten any AI

premium at all baked in but if you

actually think that AI continues to

progress in the way that it has and that

and you look at the technology and you

know it's good enough to do this stuff.

I think we're looking at a year where

those companies get some of that premium

and it's from a very from a riskreward

perspective it's pretty attractive

because they're already trading like AI

lossers what they're going to lose more

everyone is already maximally bearish

all the capital's been sucked out into

the data center beneficiaries so I think

that looking for those companies which

we did a pretty in-depth screening we

narrowed it down to 30 companies we did

some qualitative stuff too just looking

at which companies are actually already

talking about using AI for this that is

going to happen and it'll be a great

trade in 26 I think.

>> And how do you crystallize this into an

actual trade and find those companies?

Talk about the screening process you

just referenced in terms of finding

companies that have a high headcount and

you constructed a bureaucracy score and

then we have a kind of dot plot the AI

bureaucracy alpha framework.

>> Basically it's a process. First, it's

the reason why this trade kind of came

about is we were looking at we did a

very naive screen, which is just let's

take the S&P 500 and let's look at the

bottom tenth of companies that have the

lowest net income per employee. And that

those 50 companies have massively

underperformed the S&P 500. So, that was

interesting to me, but that there's a

lot of reasons that you can have low net

income per employee. So, we said, okay,

we got to go a step further. we use

first basically what you want is a

company that is spending a lot on

employment and spending a lot on

employment in an inefficient way use

SGNA as a percent of sales and then also

what you want is this margin optionality

so you want these companies to be able

to within their sector sector relative

cut the

undifferiated lowest performing labor

and be able to increase their margin

significantly because they go from the

bottom to the top of their field in

terms of margin

So we did that. We did like a

quantitative screen and then we narrowed

it down qualitatively by looking through

filtering for companies that are talking

about AI or have done headcount

reduction already. And then like a an a

really interesting aspect or really

interesting area is like insurance

brokers for example. This is like the

most paper pushy organization in the

entire world and there's a lot of

employees and there's a lot of people

and I don't want to give off the vibe

that I'm rooting for people to lose

their jobs but at the same time you have

to realize that this is going to happen

and uh when the market's giving you

these names at like incredibly

reasonable valuations you got to go for

it. So we narrowed it down to about 30

companies that in many different fields

docuine for example is one of them and

then we threw in some of the names that

would help in this transition like some

of the cloud names. It was very

interesting while we were writing this

IBM acquired Confluent. So it's it's a

it's a broad 30 names across different

sectors but the thing that they all have

in common is they can utilize AI to

increase their margins like this year

because the work that that they have a

lot of employees and a lot of those

employees are doing closer jobs.

>> Hey everyone, you heard us talk about

the Catrindex and I want to take a

moment to explain what that is. As the

name suggests, it's an index of

Catrini's most high conviction ideas at

any point in time. And the Catrinex tool

is an all-in-one dashboard for tracking

that index, as well as over 130 thematic

baskets that Catrini and the team have

made. Plus, it's updated in real time

with instant notifications as prices

move and facts change. The Catrindex is

a separate product from the Catrini

Research Substack. Up until recently, if

you wanted to subscribe to Catrini

Research and the Catrindex, you needed

to buy them separately. But now, you can

get Catrini Research and the Catrindex

together in one bundle through Substack.

Right now through January 14th, Monetary

Matters listeners can get an exclusive

25% discount on this bundle. You'll get

access to the classic Satrini Research

that we've all come to love and the

Catrindex subscription. Visit

satrin.com/mjack

to access the offer. Just make sure that

you're signed into Substack or enter

your email to unlock the special

discounted landing page. If you aren't

signed in, you can't access the special

offer. Remember that satrin

research.com/mmjack.

Let's get back into it. I'm just trying

to see the dot plot. On the xaxis, it is

companies that have a high overhead

right now and the y-axis is the

optionality to increase their

efficiency. So a company that that's a

classic example would be these

consulting companies like Accenture or

Booze Allen Hamilton. So Boo Allen

Hamilton that also has a DC factor

involved in that because everyone

thought okay Doge is going to happen and

all of these government contractors are

going to get totally destroyed. That

never really happened. But so Accenture

I know that there was a while where they

actually were seeing a very large

increase in their revenues from AI but

the stock had not been trading well at

all. what's going on at Accenture and

and you know exactly that is what what

you refer to in in the piece as a people

factory a company that's just extremely

um employeeheavy

>> and um listen you know probably both of

everyone's kind of guilty of this where

in the beginning um

all you really needed to outperform

because of AI was a conviction that we

were going to do it and then you said

okay well if we're going to do it what

do we need we need the data centers and

everything has been a derivative of that

and I did that And that's been great.

And is that going to keep happening?

Yeah, probably. I I do think that

there's something to be said about 2026

maybe being the year that algorithmic

improvements start to compete with just

like raw compute, but that's been the

trade and it's been a great trade and I

the reason why we included this specific

trade in this piece is because

that probably will continue. But at the

same time, it's sucked a lot of capital.

like it it's become such a no-brainer

and so easy that there are a lot of

companies out there that just have not

gotten a second look. Like you said, you

know, Axenture has has uh they've cut

jobs. They've spoken about AI improving

their margins. They AI keeps getting

better. I don't know if you've used

Gemini to create images, but it's like

it nails it. Like it doesn't really get

text wrong. I think you create 10, maybe

one has a slight error, and fixing it is

as easy as just saying, "Hey, can you

fix this?" And it does it. You go from

10 people each individually working on

this being managed by one guy to the

manager and the manager just manages the

AI agents. And the thing last year why

this wasn't really going to happen was

the hallucination rate was just

unacceptable. And there's still been

some enterprise resistance to that. But

you look at like Google's TPUs, right?

They're deterministic. And Gemini has a

much lower rate of hallucination because

of that. So people haven't taken a

second look at this. And that happens

all the time in markets, especially when

there's an easy way to make money that

doesn't require looking at the this

these areas. And it'll I think if you're

not looking at this, you're doing

yourself a disservice because we're not

going to spend a tr trillions of dollars

on creating the infrastructure and then

just be like, cool, done. All right,

we're spending a trillion dollars and

we're going to get to AGI and then AGI

happens and overnight everyone loses

their jobs. Okay, like that's not how

it's going to happen. It's not how

technology works. The trade here, you

know, it's not the AI takes your job

tomorrow. the the trade is companies in

2026 will slowly realize they've been

paying humans to do things that

computers are already better at and can

already do at a fraction of the cost.

And historically when that realization

spreads it's it's an interesting time

for society and for shareholders. If you

like um the push back to this is well

every time that there's a new technology

new jobs get created because of that

technology. You look at Excel for

example, it's like the it got rid of the

actuarial profession, but it pretty much

created the profession of investment

analyst. But there there's always a gap.

You don't have the lites being mad

unless people first lose their job

before those new jobs get created. And I

think that we're in that period for the

gap to to occur now.

>> And you put this trade first in your

piece. So you must have a pretty high

degree of confidence, a higher degree of

confidence I should say. Yeah. What

about this view? What about the backlash

to this uh James? Because, you know, I

actually kind of wanted to save this for

last or not talk about it because I I

wanted to avoid um you know, kind of

making people angry about everyone's

going to lose their job. It's not a it's

not a happy thought at all. Um you know,

I I think it's kind of a maybe a bad

thing, but that doesn't really matter in

the investment business, I guess.

>> Yeah. I mean, uh, like I I'm not not

rooting for people who lose their jobs,

but again, we spend a trillion dollars

on a technology and the sole purpose of

that technology is to replace people.

And as it gets better, there's a whole

other aspect of the societal

implications and could we see an economy

in 20 in 26 where the unemployment rate

continues going up, but stocks also

continue going up? That's happened

before plenty of times in history. Good

example is like after World War II, you

had all the GIS come back, the

unemployment rate was very high, but the

stock market continued to do well. Um,

it's reminis, you know, March of 2024,

we wrote a piece about how to play Trump

winning the presidential election. And

yeah, half the people that read that

were like, nope, I don't like this.

It's, you know, when when you're an

investor, you have to separate what you

want to happen versus what's in front of

your face and likely going to happen.

And it's something where the reason why

it's upfront is because there are a

couple things we could be wrong on.

Maybe the stock market continues to just

take, you know, play the easiest thing

which is Nvidia and the supply chain and

and and uh all these other things that

we talked about like advanced packaging

and the bottlenecks towards uh making

compute capable of doing AGI. But

eventually, no matter what, if you're

spending that, like this will happen.

And it's a good idea to be prepared as

an investor for what that looks like

when that does happen. And it's also got

a macro angle of just

are we going to see the unemployment

rate rise? And I can see it very clearly

of in the beginning when that starts

happening and the unemployment, it might

already be happening. The unemployment

rate is rising, people are getting

increasingly bearish, but then companies

keep posting excellent earnings like

that. That's something that's very

likely to happen and it's something that

you should be aware of. So that's why we

led with it because it's everyone I

think deep down knows that this is going

to happen. Nobody that's using AI with

any regularity doesn't witness the

improvements, doesn't notice that it's

doing more of the things that they would

be doing themselves. Um it's

controversial, but there there might

have to be with every new technology

there's a change to some degree of the

social structure and that's beyond my

IQ. don't know what the what society

looks like in that world, but it's

something that probably should be

thought about.

>> And what about the counter case that

actually AI is just a parlor trick and

it can't be used for real knowledge

work. It's just taking ideas, not

generating new ones. And that in

particular, it there's a reason that it

hasn't been picked up in the enterprise.

For individuals, sure, but for

enterprise, not so much because you

referenced the hallucination rate.

Interesting. I've always been a Gemini

user, so I've I've seen hallucinations a

few times, but not nearly as often as

people are talking about.

>> Yeah, I I think that

when you're investing in technology

changes pretty quickly. Um, and we've

seen a lot of changes this year and uh

you have to update your your priors when

when those changes occur. I would just

say, show me a time in history where we

put this much capital into something and

didn't at least make it better. We we

like like we're putting more capital in

this relative than we did to getting

humans to the moon. You can be on the

other side of that and say that it's a

parlor trick. Sure. But e it's not. But

even if it was, you throw enough money

at something and eventually it doesn't

become a parlor trick anymore. It

becomes reality. And even if the stock

market becomes disillusioned with it,

the technology will continue to go

forward. And that's what's so um

compelling about looking for

opportunities that are mispriced like

this because when you think about the

dot bubble for example, I don't

necessarily think that we're nec we're

repeating the do bubble or that we're

far along in that pathway yet. But when

you think about the.com bubble, the

biggest strides in the technology were

made while the bubble was bursting. It

concentrated capital into more efficient

uses. there was a bunch of capacity that

was there for the taking for people that

were building things. If that were to

occur, you look at how a lot of these

names find themselves in value factor or

in low volatility factor. And if you

have this top of mind and you're

monitoring it throughout the year, if

there is some sort of scare, these names

will probably go down a lot less or

maybe even go up because they're

improving their margins. So yeah, I

would say to that uh parlor trick uh

it's an uninformed take and those people

will get their day, right? They'll get

their day where they say, "I told you

look like the stock market's down. That

means the AI isn't a thing." But it will

be and it will continue to accelerate.

And if you're not preparing for that, I

it could be a rude awakening.

And so James, I'd categorize these the

basket of companies that could shed

their workforce and increase their

earnings and productivity as it because

of AI. I'd say that is a trade that is a

result of AI trade rather than a capex

trade, you know. Yeah. Um, so when you

came on my previous show in 2023 talking

about AI and everyone was a skeptic, you

were all talking about Nvidia, the

semiconductors, maybe the companies

around the semiconductors, the data

center. Now it seems you're focused more

on companies that actually are going to

use AI rather than the capex. You still

have some capex names. So, you know,

James, I I mentioned that, you know, yet

again in 2025, you know, yet another

year of the index has beaten the S&P.

However, don't you think it would have

been better to just trade the cap the

capex names? Like all these memory names

are going crazy and you probably were

long those names in 2023 and 2024, but

has the pivot from the capex into the

beneficiaries

>> is from phase one to phase two. Is that

a mistake? Might we still be in phase

one?

>> Absolutely. The if you look at the if

you go and you load up our AI basket,

it's about 30 or 35 names throughout the

year. We did own Micron, we owned SKH

Highix, we owned uh Kioia and and the

names that benefit from this increasing

need for both storage and memory. And

that basket, I don't have exact numbers,

but I'm pretty sure it was up like 70 or

80% year to date. So the yeah, it should

have been sized larger, but at the same

time, it is a the risk, right? If you

look at how it did in April, right, it

wasn't that that obviously an

opportunity to buy the dip which we did,

but there the risk it will continue to

increase and it's the

in investing in the infrastructure. Yes,

it's a it's been a great trade. It

probably will continue to be a great

trade, but it does have an expiration

date, I think, and I'm I'm not smart

enough to know whether that's this year

or five years from now. So I think you

do stay allocated to it and you do

continue to really the trade in AI

infrastructure has been much more about

monitoring for bottlenecks than it has

been about just going for whatever we

need to build because that's constantly

changing right now. uh you have uh

another idea that we speak about in this

is uh advanced packaging has become you

know the the bottleneck and and um we'll

talk about that in a second but to

answer your question yeah we probably

should have sized up a little bit more

in the in these like first phase

infrastructure buildout names in 2025

but you do need to start thinking about

a margin of safety and if that means

that you put a a portion of your I

really am a fan of like the tracker

position like same with Peter Lynch, you

know, on, you know, 1400 stocks and and

you would just buy stuff to put it on a

screen. I think it's a really good idea

to put this on your screen and then if

the tide starts changing and you see

these names outperforming that's like

like okay because this second phase will

happen and and we were definitely like

early to that but it's going to happen

like the infrastructure you every time

you have an infrastructure build

eventually you get a capacity cut and I

think we'll get that too here I don't

think that'll occur in 2026 but you

really be focused on just in the

infrastructure layer. You'd look at the

bottlenecks. Memory is a huge one.

Advanced packaging is another one. And

then I do think it's a good idea to

start broadening out a little bit in

terms of who who's going to this is a

real technology now and we're going to

keep building it, but also it's been

built to a degree that is pretty useful.

>> I think that yeah, so just looking at

the contribution of the P&L, yeah, the

dynamic AI basket was up 70% this year.

Interestingly, the biggest contributor

was shorting the VANX semiconductor ETF.

>> Yeah, that was a good one. The Deep Seek

thing.

>> Yeah. So, that's good. I don't know that

many investors who were who made money

shorting semiconductors this year.

>> I had a great conversation uh this year

with uh Peter Borish who uh was at Tutor

when like the 1987 crash and and they

they nailed it, right? and and he was

the guy that was like making the analog

between the 20s and that helped him be

prepared for that. And when I started

talking to him, I said, "Man, I just got

to ask you." And you could see the look

on his face. He's like, "This guy's

going to ask me how I predicted the 87

crash again." And I was like, "How did

you convince yourself to go long at the

at the lows?" because that is so much

more difficult if you think about like

like espec if you're right about like a

market event like that where you have a

huge draw down and you make money from

it convincing yourself to get back in is

way more difficult that that's kind of

the similar thing that happened with

where I thought that it represented this

broadening out and for a month made a

ton of money shorting SMH on that pretty

covered pretty much near the lows but

then started looking at these kind of

what about names that aren't

semiconductors and you know so so it's a

double-edged sword to to do that but I

do think the valuations are getting too

attractive and the opportunity is

getting too obvious where the profit

incentive to utilize AI is going to

become a driver and the

infrastructure I mean there's been a lot

of money spent and you start have to ask

yourself you do need to see adoption and

I do think that we will see adoption and

it will drive more money going both of

these things can do well at the same

time it just a matter of continuing to

be solely focused on the infrastructure

buildout has a lot more risk than

starting to think about two steps ahead.

But there there are places in that

infrastructure buildout that are pretty

reasonable still the advanced packaging

that we were talking about.

>> We will get to advanced packaging. I

just want to remind our viewers that I'm

looking at the attribution of all of

these baskets and the catindex that is

available in a package deal. the

Catrindex and the Catrini Substack in

one bundle for 25% off to Monetary

Matters listeners until January 14th.

James, talk to me about advanced

packaging. I literally before I read

this, I thought you were talking about

some sort of really fancy box for a VPS

or a FedEx. So, I'm I'm such a lit, but

what kind of companies are we talking

about here? What's the theme? Why is it

so important for custom silicon as well

as maybe Nvidia too?

>> That's the advanced packaging. It is

what it says, right? It's packaging

these chips in a way that that basically

for 50 years, um,

you know, if you're familiar with

Moore's law, we made transistors

smaller, and that worked great. And then

we hit a wall where chips like like the

we could make them bigger to fit all

this stuff, but we can't make them any

bigger. They won't fit on the lith

lithography machines. It's called the

reticle limit. So, if you try to print a

chip bigger than that, you know, the

yields collapse, you lose money. So the

new game isn't make the chip bigger.

It's make a bunch of small chips which

are adorably called chiplets and you

know stitch them together so they act

one big chip. And that stitching is

advanced packaging. And you know um it's

it's this is an interesting bottleneck

for me because it's been a bottleneck

for a while. If you've heard about KOS

from TSMC that's been a big capacity

constraint in making as many GPUs as

possible. But the reason why this is

interesting to me now with names like AM

Core and and some of the supply chain

like the tooling and stuff like that

we saw Google come out with these TPU or

they've been doing the TPU but the TPUs

got good enough to deliver a model that

was state-of-the-art and

that's not going to be the last piece of

custom silicon that we see if Google

makes their own you know is is competing

with Nvidia with TPUs or if Meta makes

their own chip or if Nvidia launches um

Blackwell all of these need advanced

packaging and they're all fighting for

the same exact capacity. TSMC is the

only one doing it at scale and they're

they are completely tapped out. And the

names that are taking their overflow

capacity like ASSE technology, they

trade at a much more reasonable

valuation than say Nvidia. And

there's another name that might be

controversial or but Intel, right?

Intel's been dead for a while. They keep

screwing up. Intel's foundry business

might be a mess, but their packaging

technology which um you know eBI which

is there's basically if you think about

packaging a bunch of small chips

together, there's only so many ways to

do it, right? And because there's only

so many ways to stack stuff. You can

either do it in two dimensions where you

just put it on the chip and then drill

through it or you can do it in three

dimensions where you stack things on top

of each other and then stack them on top

of the chip and it looks like a cube.

their advanced packaging. E-IB is their

2.5 dimension which is just two

dimension basically. It's a marketing

thing and FOS is three dimension. It's

becoming the first kind of real relief

valve for this massive bottleneck.

There's already rumors that Apple and

the hyperscalers are looking at Intel

just for this packaging layer. So the

trade for advanced packaging has these

tailwinds from

on one hand you get upside to this

increase companies don't want to pay the

Nvidia tax anymore so they develop their

own custom silicon and then at the same

time it's additive so if Nvidia sells

more chips this advanced packaging still

this complex still does well and so like

amcore it's a boring steady has some

exposure to mobile it's increasingly

picking up the volumes that int or TSMC

are too busy to handle. And then if you

really don't want to own Intel, you

could buy the guys that are like selling

them, the staplers, a company called

Kulic and Sofa. Uh I don't think I'm

pronouncing that, but the tickers click

tail. I see. They make this specialized

kind of tools that that bond these

chiplets together. I think everyone kind

of frames this as a binary thing. Either

Google wins or Nvidia wins or Chinese

ASIC wins or it right now at least it's

all additive. So I think there betting

on like the duct tape holding the chip

together rather than who wins the chip

war is probably a good idea here. So

we've got a chart and I like how you

make this very simple actually it's the

colors of my high school so I like it

even more but so that over time the

share of incremental performance from

chips mostly it came from in the old

days it came from the orange the

transistor scaler so basically the chip

being more efficient and now it's coming

from the blue the system level

integration the chiplets the high

bandwidth memory HBM and the co-acked

ASEL so it's not about fitting as many

little tiny little things on the chip

anymore is increasingly becoming about

connecting the chick the chip within the

ecosystem. Is that what's called what

internet interconnect means?

>> Yeah, you need Yeah, you need

interconnect. So, basically, you need

all these chiplets that talk to each

other, you know. So, so yeah, you and

and uh that's um yeah, it's kind of

interesting once you start like uh like

read you're like oh advanced packaging

it's packaging stuff together advancedly

like you know it's like oh interconnects

it's connecting stuff inter

you know the the obvious um obviously if

you want to be an investor in this stuff

unless you want to go the route of like

literally becoming a semiconductor

designer so that you can understand this

that's like more power to I would much

rather understand it from speak to

people that are super smart and then try

to make it understandable to someone

that's not like myself.

>> And so in terms of the core companies

for this advanced packaging, the core

longs you call them Intel, Amcor,

Synopsis, KIC, and BESI. Tell us about

Synopsis. Later on in the 26 piece, you

wrote a one name piece just on Synopsis

and then also BESI that that company.

So, Synopsis basically Synopsis is

another winner of this custom silicon

drive, this aversion to the Nvidia tax.

This idea that companies can only

maintain massive gross margins for so

long. And it's got an interesting story

because

it was punished for a pretty long time

on the Intel. But the elevator pitch is

basically you can't like like this is

called EDA and IP. So basically they own

the IP of chip design and you can't

build a modern chip without them. It

costs like $750 million to design a

leading edge chip now. You're not

drawing it on a napkin. You need to you

need this EDA electronic design

automation software to simulate every

electron before you spend even a dollar

manufacturing it. And this is

essentially

duopoly or triopy. You have a synopsis

cadence and mentor which is owned by

Seammens. And it it the synopsis

specifically has gotten beaten down

because they had a massive contract to

help Intel port designs to their um 18A

node. Uh Intel cleaned house. They

brought in Lip Bhutan to look at the

books. He realized that Intel was

basically paying for an ecosystem that

didn't exist yet and just rug pulled the

contract. Synopsis had to eat the loss

because you can't upset your one of your

biggest customers. The stock went down

like 40%. and the market started pricing

it like it's this structural flaw, but

it's a contract dispute. Intel's still a

huge customer. Intel is improving.

Intel's going to need more EDA and IP.

They'll eventually pay back those fees

because they need Synopsis to make their

18A node work. And if you look at the

valuation gap between Cadence and

Synopsis, Cadence is like a 45 times

earnings. Synopsis is at 30. Synopsis

also did a great acquisition with ANCIS

which is like the physics simulator with

so it's it which is benefiting from

these advancements in AI and also

enabling these advancements in AI which

I think is pretty poetic and like we

talked about earlier that this advanced

packaging 2D or 3D like as we move

towards three-dimensional making these

cubes of chiplets you can't just code

the logic anymore you have to simulate

the physics so heat dissipation warping

melting

By owning Synopsis and Synopsis owning

ANCIS, they're positioned themselves to

become the physics engine and you're

basically buying what's what can really

be thought of as a monopoly at a huge

discount because of a like breakup. Is

it like a little bit expensive? Yeah,

Tra software multiple asset light and

it's a huge discount to Kanan. So, I

think the synopsis is a great play for

2026.

>> So, that is a play for the packagings of

the actual chip. The trade idea right

after that is something that got me

really excited because early on you had

this infographics about a picks and

shovels play on AI and whenever you

talked about a picks and shovels play on

AI uh it it tends to have worked out

well. This is something James I've

actually been looking at myself. I've

just been asking Gemini what are some

commodity materials that are making uh

AI? You know it turns out silicon is

extremely available tight. There's not

going to be kind of a shortage of the

actual sand. There's a lot of sand on on

Earth. Um, but what could there be a

potential shortage of? And what could

there be a potential buying in the cycle

where the stocks aren't priced have no

AI premium at all and yet they are

producing a commodity or a material or a

service that is not that is going to be

in great demand for this AI buildout.

>> So this is like this gets like way more

nerdy. This is like the trade here is

basically you have a bunch of companies

that are essentially commodity companies

that are putting the kind of materials

into if you think about the GPU is like

the brain. These guys are like building

the skull and the spine and everything

that goes into it. And a lot of these

bottlenecks that get discussed, they're

digital scaling running into physics.

And if you look at the way that this has

progressed, we seemingly always have a

bottleneck somewhere, whether it's like

in in a certain resin or a type of glass

or so a lot of these companies have done

really well. If you you can split them

up kind of like oil and gas. You have

upstream, midstream, downstream, and

downstream ones like the PCB names like

Celestica, which make the the boards to

put these chips onto, they've done

great. And then midstream's done a

little less great and upstream's done

not so hot. PCB stands for printed

circuit board and yeah, Celestial was a

stock that you had such high conviction

and you actually wrote a single name uh

piece on it and we're very long it.

>> Yeah, that was a great one. It's 17

bucks. I think it's 300 something now. I

sold it already but god that was a

mistake. But so looking at some of these

like chemical or like material

companies,

if you look at the media out of like

Japan and Taiwan, they're all talking

about these these these uh shortages in,

you know, not to like bid like BT resin,

T glass, the like non-conductive film

probably the and it's all stuff where

like nonconductive film for example,

when you stack those high bandwidth

memory chips that that you need for any

AI accelerator, you got to glue them

together and and you got to gloom with

something that doesn't conduct

electricity but conducts heat. There's a

company called Resinac that has like

100% market share on that film. And the

you look at and when these companies

which are priced like chemical or gas

companies experience this positioning as

like the only thing that we need more

of, they do tend to increase capacity,

but their stock also goes up 350%. Look

at Nitobbo in Japan. They make they make

tea glass fiber which is a special type

of glass that doesn't expand when it

gets hot and they have a 70% market

share. You you can't build a high-end a

AI server board without this T glass and

that stock has absolutely killed it. So

it's interesting to look at that and say

where else might there be bottlenecks?

Maybe these bottlenecks continue. So we

created like a watch list of

categorizations. Here's you need this BT

resin, you need nonconductive film, you

need this, that and the other thing. And

some of them won't have bottlenecks.

Some of them are very commoditized and

it's easy for competitors to to increase

capacity. Some of them will and watching

that in 2026 will be increasingly

important. An interesting one that I

liked a lot, you you remember MSG,

>> not Madison Square Garden, but like when

you order Chinese food in like the early

2000s, you're like no and no MSG please.

Yeah. So there's a company called

Ainamoto which makes MSG

and they also make the quoteunquote

buildup film ABF substrate if you want

to be like a huge nerd and that's like

the insulation layer. They have a 90%

global market share. So it's and guess

what they trade like they make MSG. So

it it they're boring. They're buried in

the chemical sector. And if you look at

this supply chain, the closer you get to

the AI box, the box that you can touch,

the better the stock is done. Look at

like outside of the actual GPU, you look

like Verdib or Victory Giant, like you

said, Celestica, TTMI, the Serverax,

SMCI for a little bit, not anymore. as

you go upstream from that, the guys that

are like mixing the resin or there's

still a lot of them that are trading at

these cyclical chemical multiples and

you have a really good proof of concept

in something like Nito which traded like

that and then everyone said whoa we need

a lot of teal glass and boom done. So I

think the trade here is much more

interesting like the midstream maybe the

upstream if you want to be like if you

want to play the upstream you got to be

really on top of monitoring these

bottlenecks but the midstream is

benefiting from both sides just keeping

like a news alert on like country Taiwan

or Japan word shortage like that that I

think that that and then you know once

you hit one of those you don't have to

be like a again you don't be a semi

engineer you just go load up 26 trades

and you hit one and it comes through and

it says hey tan Telm there's a tantelum

shortage. Okay. Control F tantelum and

it looks it'll show you know. Okay. So

tantelum capacitors are made by Marada

manufacturing. Okay cool done. that

that's it's something this happened to

Marada in I think it was 2021 too where

they said we have all these crypto

mining GPUs and a crypto and these

crypto mining A6 and these they use 10

or 100 I don't I might be off an order

of magnitude as many times of these

capacitors than normal chips and it was

the most severe shortage because of

crypto mining that we've ever had

looking at something like that and then

if they start talking about ah we need

these guys to make more of them you find

a company that has 170 to 100% market

share and that can go on for a lot

longer in this environment than people

think.

>> And so you're looking at companies like

yeahbo but in particular Resinac that

seems to be the company that you had the

most high conviction in.

>> So yeah basically Resinac is has the

most shots on goal here where Resinac

has a lot of areas that go into the AI

supply chain. Not there aren't any that

are like

really in severe shortages yet, but

basically if there's going to be a

shortage like in the midstream eruption,

it it Resinac has the highest kind of

likelihood having a a significant market

share in that area. So again, it comes

down to this is a watch list. You can

play the existing shortages, but I would

just warn like these companies still are

commoditized. They still make they're

companies that are making things.

they're turning raw materials into stuff

and the companies are some of them

selling raw materials like if that

bottleneck gets resolved these companies

going to have 50%. You know so so it's m

I think it this is a trade where you can

put it on and take the risk of I think

that these architectures won't change

that much and we're going to keep

needing whatever it is whether like te

glass will probably need more of it.

We're also like you could look at like

glass substrates and stuff like that,

but you could also just say this is a

great watch list and I'm going to just

wait until because the best way to be

early to a trade is to just actually be

paying attention to something. just

paying attention good example like it's

2022 chat GPT you say okay I'm going to

pay attention to what it takes to run

this model or to train this model or you

know whatever to deliver this product

and you say okay it's GPUs and then you

start seeing everybody talking about

buying GPUs okay I'm going to buy Nvidia

it's but I and that was an easy one and

now that's the most obvious thing in the

world you got to start saying okay well

what don't what what is what are people

not really paying attention do. And for

me at least, like before I did this

research, I didn't really know about

buildup film or ABF substrates. And

that's fair, but I'm sure there are a

lot of people that are listening that

might know about that. They might be

engineers, but for me, it's it's easy

enough to just monitor these shortages.

And once there is one, you buy the

companies. And again, if you look at

Intobo, it's been going up every single

day for eight months. And the fir you

look at the first headline about the

shortage of T-class and it's eight

months ago. So

>> you got to track the shortages

in terms of actual demand for commodity

things like copper, maybe silver, but

also natural gas. Where do you think the

biggest potential for the price going up

could be? I know that you're getting

increasingly interested in the natural

gas area and in in your model portfolio.

This is Trendex that actually is a is a

major theme. Why have you gotten so

bullish on natural gas, a commodity that

is famous for the supply being able to

just go up a lot?

>> Yeah, the so we're bullish on in the

commodity space. We're bullish on copper

and natural gas. Copper is a little bit

of a simpler story, but they both have

the same kind of thing where you don't

necessarily have this belief that this

cycle could result in in any sort of

sort of kind of competition or any sort

of like struggling to get supply online.

And the thing about the copper might be

the easier long natural gas and we do

own a bunch of copper miners and we have

for a while on that thesis but natural

gas is much more interesting to me

because it's like nobody believes it and

I get it and you can can look at we

published our thing on natural gas in

September. Natural gas went from sub $3

to like above five and then it got cut

in half and it's yeah like natural gas

trades on the weather because it's the

only thing that people care about in in

the front month. The opportunity here is

basically if you have the if you believe

in the idea that like this LG export

capacity will come online and it'll

start competing with the data centers

that like nuclear is great, solar's

great. I get it. We should probably

we're going to use that too, but we got

to build machine God now. Most of these

things are being run on natural gas and

we're building so many of them and it's

going to start having a competition with

this other huge mega trend in

infrastructure construction which is LG

export terminals and the US used to

import LG and now we're going to become

or we are the largest exporter. So I

think that the trade again the trade is

not I got a lot of people when we wrote

it to calls on like comtock which like

great trade okay that that worked for a

few months and then the trade is you own

these companies like EQ or comtock or

and with the idea that like the back end

of the curve will still be super

volatile because it's natural gas that's

just how the commodity works but that

volatility will trend upwards because

you'll need like the the excuse of the

perian will just bring on capacity

forever. I think like that's definitely

how it's priced and you don't have a ton

of downside from that being the case,

but you do have a lot of upside from

that starting to get challenged. I think

probably

in the next year

>> if you were you you remember like the

IPS like Vistra and so the big thing

that happened with them that was like so

amazing and got all these growth

investors to start putting a growth

multiple on it was

>> they started offering power like fixed

price contract to hyperscalers.

You're talking about independent power

producers like Vista, Constellation,

which unlike regulated utilities where,

oh, I'm going to the government to get

my 5% price increase approved, like they

can just charge whatever they want and

they're not very regulated. And you were

early in talking about those and

investing in those. But yeah.

>> Yeah. And and you and you see like what

you know like like something that's

basically a utility in early 24 start

trading within a beta of one to Nvidia.

And that's cool like the because that's

a massive rerating and really the

driving factor there was vis like if

you're a hyperscaler you're worried

about power cost you're worried about

power volatility and vistra is worried

about the same thing and vis says okay

we'll offer you a fixed price contract

on power for the next x amount and what

that enables is growth investors then

can say okay I'm not really taking a

view on like the commodity pricing of

power I'm taking a view on like

hyperscalers using more of it and I can

put a multiple on that because there's

this and I think that in 2026 well we'll

see those hyperscalers will probably

negotiate fixed price contracts for

natural gas with like like EQ has

already talked about this a little bit

it hasn't happened yet but once that

happens the natural gas equities will I

think trade better and then so it's the

trade is basically one natural gas is

going to keep powering our efforts

towards AGI until until we can do cult

fusion. And then on the other hand, the

natural gas producers are going to be

able to negotiate things with

hyperscalers that allow their investor

base to broaden out and get more of this

growth investor in there. Um, it's a

similar but not identical story for

copper, which most of the guys that are

like trading copper futures, they're not

going to believe this like AI super

cycle bull case, but we have a we have a

line in there about like how much it

would cost to build a new escandida and

when it would come online and it's

massive and it takes forever.

>> It's a giant copper mine. Yeah.

>> Yeah. Yeah. And and um so that's another

those two commodities are interesting to

me. Copper obviously has been the easier

trade. Natural gas will kick you in the

nuts a few times, but I think it's also

a similar opportunity of brightening

this kind of secular trend.

>> Have you looked into silver using AI?

>> That's I missed silver. Alex Campbell

was talking about it for a while and

absolutely nailed it. And I had that

like weird psychological like anchoring

bias that you get sometimes where I had

a whole article ready about the gold

miners and then I decided not to publish

it. I decided not to put the trade on

and I really should have and it soured

me on precious metals and I just missed

the silver trade entirely. I would say

if you are interested in that, you

should definitely go read what Alex

Campbell has written about because he's

been 100% right and I totally missed it.

>> That's nice. So that's the macro thesis

for natural gas. I'm looking at

Citrindex. You have all of these names

in your natural gas basket. You one of

them I will say is uh Texas Pacific Land

Corp, a company that I just think is

fantastic. But why so diversified? If

the core exposure for most of these

companies is the same, what is the

advantage of natural gas producer A

versus natural gas producer B?

>> It's interesting. So the biggest weights

are you know comtock and e and then but

there's a lot of really interesting

opportunities in like Canadian natural

gas which has been a nightmare and

that's so it's very much an approach of

casting a wide net because this would be

such like this thesis working out which

again this is like a longer term thesis

right it's great that natural gas went

up a ton after we published and that the

winter was cold our thesis is not that

the winters are getting cold that's if

anything like the thesis was like people

have gotten a little complacent about

the winter being cold and Sometimes the

winter is cold, but it's such a it would

be such a shift for the market to not be

priced like the perian can just infinity

ramp to meet all these demands for

natural gas that and then in the

meantime like if that isn't the case

it's the downside like relatively

limited especially from where we got in.

So it's something where it's like a

broader net of you have the upstream

comtock and EQT and then you also have

some of the Canadian ones which like are

super cheap and and whichever one and

that's a political aspect too but

whichever ones start doing well those

are the ones I add into

and like the and it's it's a fixed pie

right so it's not like we're like

marting to natural gas but the equity is

doing increasing it's also this is like

even more heretical than talking about

natural gas, but I do own a couple of

regulated utilities as well, which is

like insane. For a long time, I never

touched like my kind of framework was

like financials and utilities, they're

these industries where everything's okay

and then one day you wake up and

everything's really not okay. But some

of them really, if you look at Excel,

for example, I had a discussion with a

company that's working with them called

form. they're private and the concept of

there's been real really no incentive to

allow these utilities to increase their

capacity or to do capex spending and and

you know and with the data center stuff

like there's going to be backlash and

there's going to be a real kind of drive

to to to change the political landscape

to get them to be able to do that and to

sell more so that prices don't go up on

like households. So it's that's a super

it's a very left tail or right tail

event type thing but I do think that it

could change a little bit in in 26.

>> So talking about a potential supply

squeeze in because of AI demand it's

going to see high AI demand is going to

be so high. So copper, silver, natural

gas. What about this thing you call What

about this thing you call post-traumatic

supply disorder PTSD where these

cyclical markets have suffered for a

long time of building up inventory and

only to lead to a cyclical crash that

they are avoiding investing in the

capital c expenditure. The most obvious

theme right now would be natural gas

turbines. So like GE Vernova or Zemens

talk about that. Yeah.

So

basically last year when we wrote 25

trades, the second trade was Crouching

Tigers, Hidden Dragons, which basically

the thesis was started from the insight

of what can we extrapolate from the

names that have done really well

into insight about what's going to do

well next year. And we looked at like

Carvana and Apploven and Unity. And the

kind of common thread between all these

was which we talked about on our last

podcast about 25 trades was that they

had done so much in terms of they had

built so much infrastructure not like

necessarily physical infrastructure but

they had used the zero interest rate

policy environment to create this moat

that would be incredibly difficult for

new challengers to to come up against in

environment where interest rates are 5%

and at the same time they'd gone down

90% from their Horizon 21 and lost their

investor base and everyone viewed them

as like a bad word and nobody really

wanted to own them. So it was hated and

but also benefiting from this bunch of

cheap money that had they had used to

build great infrastructure and that did

very well in 25. So we went back to the

well of inspiration to think about okay

what else did what did really well in

2026 and was there any common thread

between them. So you've got you have

memory things on Micron SKH highinex and

NAND like kioski kiosia sandis western

ditch you've got the gas turbine

companies and

both of those sectors absolutely

murdered it this year and when you look

at them you

>> some of the best companies in the world

in terms of stock performance you know

>> yeah and you look at them you say you

can take the easy way out and say okay

AI exposure but there's a lot of

companies that have AI exposure that

didn't end up being the best performing

companies in the of the year. So

I think that if you think about it like

what is the kind of commonality between

these these companies had over the past

5 to 10 years they had these big cycles

where they ramped capacity into it to

meet what they thought was like secular

demand and then they got absolutely

screwed by doing that and now they're

reticent. they're not as gung-ho and

eventually they will increase capacity

but at least for the past year they were

pretty much content to let their

backlogs grow. average selling prices go

up. And so it's like PTSD, right? Like

once bitten, twice shy. So we call it

post-traumatic supply disorder, which is

they listen to a bullish forecast, they

built a new factory or they built a new

whatever a unit of capacity, they spent

billions, demand fell off a cliff, and

now they're just, okay, we're not going

to do that again. And it's been a pretty

volatile environment for cyclicals over

the past five years. So these like

wounds are fresh. And in both those

areas, demand came back for things like

data center power or high bandwidth

memory or DRAM or NAND. And in a normal

kind of textbook world, these companies

rush to build new factories to capture

that market share from that demand, but

they aren't. That yes, like right now,

some of them are planning on increasing

capacity. But look at Genova, right?

They're projecting their EVA margins to

go to 20% over the next years from 14.

And because

>> Yeah.

And it's they're because they're like

hypervigilant. They're not believing the

hockey stick charts. They're hoarding

their backlog. They're like they're

treating debt like a STD that they got

in college. Like the and

it's interesting say, okay, where else

could this play out? And the screen

basically that we use is okay, first you

need this trauma to exist. Then you need

then you need there to be some sort of

increased demand. And

Then you need there to be some sort of

capital discipline and also you need it

to be an igopoly, right? So one that

like meets the first three is any of the

gold miners, right? That like they have

the first three but they're not but like

the capacity can get brought on by any

of their competitors, right? It's very

it's not a concentrated market. So

some areas that are interesting to look

at from this perspective like solar for

example an analog semis wind turbines

offshore drillers like these all had

like there each one of those things that

I just said there's a very good case for

like why they shouldn't do well but if

>> low margins high capex and very cyclical

and a tendency of the CEOs in the

business to be wildly optimistic and the

stereotype of the oil CEO who just loves

to drill drill holes and borrow as much

money to drill as many holes as

possible.

>> Yeah, exactly. And and you know and and

then uh but you have you know like for

solar uh you got like uh China's doing

anti-involution which is like like uh

they're like hey guys let's stop doing

this like race to the bottom where we

just try to flood the market with as

much capacity as possible. There's a

national security element where they're

not the US isn't necessarily buying

Chinese made solar and you've also you

got like the beginnings of like green

shoes for demand with like you are using

solar as well for AI data centers and

rates are going down so maybe we see

some return with residential and yeah

then you look at a company like first

solar and if this de facto becomes a

igopoly because we're restricted from

buying in China and then the demand goes

up and They don't they're so burned by

what happened before that they don't

immediately increase capacity. That

could be really good for their average

selling prices and it could be really

good for their stock. It's and it's an

area that's similar. I mean it. So, like

I said, the gold miners don't

necessarily fit this because even though

they have demand, they're not being

relatively disciplined. They're not

agobling. But if you look at lithium,

it's still commodity. They're not

necessarily price they're not price

makers. They're taking price from the

commodity set. But it is pretty

concentrated, right? Like the amount of

players in this for mining at least it's

pretty concentrated and all of them have

gone through this trauma. So it like

something like SQM or or you know um

Alra which used to be Pedmont like like

the there is a potential there for the

same dynamic to play out. We basically

created a screen and then we split into

four sectors pretty much represented by

analog semis solar there's some offshore

drillers in there. There's some lithium

and there there's some oneoff companies

like like in like probe cards and stuff

that that aren't necessarily

>> what's that probe what

>> probe cards? Yeah, techno probe in in

Italy. Like there are just one-offs from

like individual sectors that meet all

these criteria and we did like I think

that we will see at least one or two

areas where this plays out again in in

2026. And I also think that the areas

where it did play out in 2025 continue

and it's probably a good idea to watch

for when that capacity gets really

ramped up again the when the SKH and

Micron start acting like the oil company

CEOs just drill baby drill make as much

as we possibly can that'll be probably

closer to the end than the beginning but

we're in gas turbines too we're not

seeing it from Seammens or G Vernova

they're not ramping that capacity they

have demand that's projected to double a

decade ago this company would have said

Okay, three new factories. Let's go.

They and today they're like, we're going

to do like a tiny cautious capacity bump

and then we're going to raise our margin

targets massively. And they're just

scared to death of overbuilding. And

that's great for the stock.

>> And so you had a framework of three

things, whether it's igopoly score, it's

demand score, and its discipline score.

And we can put up right now the dot plot

we'll call it of on the x-axis the

degree to which it's a price maker and

then on the y- axis the degree to which

it has capacity restraints. So for

example like pneumont a gold miner has

very high on both but it's not an

igopoly. You point out that quite

correctly the precious metals miner is

quite diversified. I didn't know until

reading your piece that lithium was very

concentrated in terms of production. I

have PTSD, not so much for investing

myself, but just doing interviews on

lithium where it was just pitched as

this is like digital gold and it's

powering the thing. And then I probably

did those interviews at the peak in

lithium prices and you know, I mean,

people were pitching um that Piedmont

stock that changed its name. There's a

reason it had to change its name like

99%. So, I just feel kind of uh

cautious, but maybe my caution is is the

post-traumatic supply disorder that

you're talking about. So, maybe that

kind of proves the point. Yeah. And you

know that it's the same thing like you

could have just as easily have done that

interview maybe you did I don't know but

about Carvana in 2021 and look at how

that did. It's like like the so um yeah

I guess you um always starting from a

place of what was hated this year and

then started ripping and what would it

take for it to continue going up and

then what else is similar to this that

that could if one thing changes that I

can track could experience the same kind

of dynamic.

Oh, James, now I want to talk about it's

not a 26 trades, but it is an idea, a

debate that is animating the market

right now, which is the degree to which

Google can beat everyone else just by

building its own TPUs, tensor processing

units. So, it doesn't need GPUs or the

CPUs. It's building its own

infrastructure. Maybe Google is going to

be selling TPUs to external parties. And

also Google with its Gemini can beat

open AAI and basically the business

model of investing hundreds of billions

of dollars in building out capacity and

buying NVIDIA chips. You don't need to

do that because Google can be way more

cost- effective. In other words, we've

all seen the famous chart from KU of how

Google ecosystem is beating out open AI

ecosystem. Not that open AI is publicly

traded. Microsoft has exposure to open

AI. So like Google has been crushing

Microsoft just in terms of share

performance. The other dichotomy is the

degree to which companies can be need to

buy Nvidia chips or AMD chips versus

custom A6 and you've been talking about

this for two years but I guess now it's

really at the fore but just talking

about that first topic like you had a

piece on the Catrini Substack about I

think it was called carving out the TPU

really good piece your thoughts here on

where are you on this debate

>> again it's something where the market

views it very much as binary it's like

you're either bullish on Nvidia you're

bullish on Google

We were bullish on Google because it was

treated as an AI loser when it was very

obviously a winner and then a couple

things went right along that timeline

and it did really well and we will

continue to see this happen. We will

continue to see the like reluctance to

pay the NVIDIA tax leading to

hyperscalers making custom silicon. And

the other like very interesting area

here that I didn't talk about that much

in the piece I talked about a little bit

but

there's like this I would call like the

silicon curtain right if you look at

like for a long time the anticipation of

people who were bearish on Nvidia was

that the custom ASIC would come out of

China who would immediately kind of

commoditize the space and that hasn't

happened yet but paying attention to the

bottlenecks of like like China is

investing a ton of money in trying to

create its own AI accelerators and

they're not there yet. But investing

along those bottlenecks is a great it's

been a great trade and it's interesting

to look at when that became a great

trade. If you look at like when the

Chinese kind of semiconductor complex

started really outperforming, it

happened right around the rare earth

export restrictions and it's something

where like the the it really this

outperformance in the Chinese AI

semiconductor complex starts happening

significantly after the July 15th export

restrictions like the rare earths and

then it goes really parabolic when some

of the loopholes for the like foreign

own fabs in China get closed and it's