LongCut logo

Steve Eisman's 2025 Wrap Up: 4 Themes Defining the Market Today | The Weekly Wrap

By Steve Eisman

Summary

Topics Covered

  • AI Growth Masks K-Shaped Economy
  • Power Constrains AI Data Centers
  • Scaling LLMs Hits Diminishing Returns
  • Private Credit Hides Fraud Risks
  • Housing Unaffordability Demands Supply Reform

Full Transcript

From the bottom of April, the market is up an incredible 37%. NASDAQ swings have been even larger. It's been quite a year. In this wrap, I'm going to analyze

year. In this wrap, I'm going to analyze the year, its trends, and themes.

Although inflation has slowed, the cumulative impact of all that inflation over the last several years continues to be felt. The private credit debt

be felt. The private credit debt universe has ballooned from around 2 trillion in 2020 to roughly an incredible 3 trillion in 2025. Investors

are worried about the growth in private credit and what potential losses lie underneath. In my view, the potential

underneath. In my view, the potential realization that the ever larger LLMs won't get us to the finish line is the fundamental risk to the entire AI story.

AI has been the tide lifting many boats and the reverse would not be pretty.

Just like in the financial crisis, the signs exist if you know where to look.

Hi, this is Steve Eisman and welcome to the weekly rap and year-end rap. In this

rap, I'm going to analyze the year, its trends and themes. This will be the last wrap for the year. The next wrap will occur on Friday, January 9th. First, I

want to thank all of you viewers and listeners for tuning in and for sending in your questions and comments. I know

I've learned a lot from your comments, so please keep them coming. And now for the year. The data we will discuss will

the year. The data we will discuss will be as of the close on Friday, December 12th, 2025. For the year, the S&P 500

12th, 2025. For the year, the S&P 500 index is up 16% and NASDAQ is up 20%.

It's been quite a year. The year started quietly with most investors confident that President Trump's policies would be promarket. By February 19th, the market

promarket. By February 19th, the market was up a bit over 1%. And then the tone changed as President Trump began to

unveil what he had promised during the entire campaign. He was going to impose

entire campaign. He was going to impose tariffs on every country that trades with the United States. President Trump

unveiled his tariff plans on April 2nd, Liberation Day. From February 19th

Liberation Day. From February 19th through April 8th, the market declined 19%. And then, as investors realize that

19%. And then, as investors realize that maybe the tariffs would not be so bad, the market rallied for the remainder of the year, led by the AI story. As of

Friday, December 12th, the S&P 500 is up 16% for the year. But from the bottom of April, the market is up an incredible 37%. NASDAQ swings have been even

37%. NASDAQ swings have been even larger. It's been quite a year. Now,

larger. It's been quite a year. Now,

let's review some of the big themes of the year and then dig down into sector performance and how these themes reveal themselves within the 11 sectors of the

S&P. I think there are four dominant

S&P. I think there are four dominant themes in 2025. One, the K-shaped economy. two, the AI revolution or an AI

economy. two, the AI revolution or an AI bubble. Three, fears about the growth in

bubble. Three, fears about the growth in private credit. And four, affordability.

private credit. And four, affordability.

Theme one, K-shaped economy. Despite the

economic uncertainty created by tariff news, US GDP should grow over 2% in 2025, which is pretty decent. But under

the hood, there are emerging problems. 2% of 2024 US GDP is around $500 billion. If you add up all the AI

billion. If you add up all the AI related capex for 2025, it's over $400 billion. So outside of AI related stuff,

billion. So outside of AI related stuff, the US economy is barely growing. It's a

K-shaped economy and that K is reflected in the performance of individual stocks and sectors of the market. Specifically,

despite a great overall year, 37% of the stocks in the S&P are down for the year.

This is also why the S&P 500 equal weight index is up only 10% versus the 16% for the S&P 500, which is a weighted

average index. The big got bigger. Theme

average index. The big got bigger. Theme

two, AI revolution or AI bubble. AI

related stocks and themes drove the market higher since April 8th. Stories

about Nvidia's enormous revenue growth and hyperscaler capex dominated the market. Yet, even though the market is

market. Yet, even though the market is either at or close to all-time highs, it's essentially been flat since late October. The market has been wrestling

October. The market has been wrestling with whether there is an AI bubble. Let

me address the bubble debate. One of the great things about doing this weekly rap is that I get to revisit themes and refine my thinking. I've given this

bubble debate a great deal of thought.

And here is what I think are the major issues. As for whether it's a bubble,

issues. As for whether it's a bubble, there is no question hyperscalers are buying chips like crazy from Nvidia, AMD, etc. Because the hyperscalers have

actual demand from customers. AI data

centers are being planned and constructed because there is real demand. None of this is imaginary. This

demand. None of this is imaginary. This

is not like the dot bubble where hundreds and hundreds of companies went public with little more than a business plan. The largest companies in the

plan. The largest companies in the world Amazon Google Meta Oracle are spending billions of dollars in capex

because they have actual orders. Right

now, there is a virtuous cycle because the LLM models that create AI continue to scale ever larger. And as they scale,

they have an insatiable demand and need for more and more chips. So the demand is real and the money is being spent by

large global companies. The fact that Nvidia's third quarter 25 revenue grew in excess of an amazing 60% is really all you need to know. The bulls argue

that it is still very early in the AI story. For now, we should definitely

story. For now, we should definitely expect the AI growth story to continue well into 2026. But what can potentially

derail it? I think two things. One

derail it? I think two things. One

structural and one intellectual. On the

structural side, the current binding constraint on AI growth is power. You

can build an AI data center, but if you can't plug it in, what good is it? And

that has already happened in Santa Clara, California, where two finished data centers have no electricity. The US

produces currently about 4,200 terowatt hours of electricity annually.

US electrical growth used to be 1% annually. It is now 3%. Now 3% does not

annually. It is now 3%. Now 3% does not sound like that much more than 1%. But

3% of 4,200 terowatt hours equals 125 terowatt hours. And that's the equivalent of

hours. And that's the equivalent of these power needed to supply a few large cities. New York City, for example,

cities. New York City, for example, consumes about 65 terowatt hours annually. This is why hyperscalers are

annually. This is why hyperscalers are looking for alternative power sources.

It explains the incredible performance of nuclear and uranium stocks and other alternative energy stocks like Olo and Bloom Energy. Yet, here's the problem.

Bloom Energy. Yet, here's the problem.

It takes time to construct new power sources. It requires planning,

sources. It requires planning, permitting, and construction. And that

all takes lots of time. From drawing up the plans to completion, it's at least a 2 to threeyear process. Pray for mild winters because otherwise too much

stress will be put on the grid. More

importantly, if it becomes clear that AI database construction will be delayed because of lack of power, hyperscalers will slow their ordering of chips and

the entire virtuous cycle will reverse.

This first risk that I've just outlined is a question of timing, but does not undercut the overall AI thesis. The

second risk is an intellectual one and goes to the heart of the entire enterprise. Is the approach to

enterprise. Is the approach to everinccreasing LLMs a dead end? Back in

the great financial crisis era, when my team and I researched the subprime market, we eventually concluded that the entire fixed income mortgage industry

rested on the bedrock of one assumption and one assumption only. Housing prices

in the US could never go down on a national level. Once we realized that

national level. Once we realized that because of bad underwriting, housing prices were in fact going to decline, we also realized that the entire mortgage

market edifice was going to collapse.

The fundamental risk to the entire AI story wields potentially similar dynamics. The bedrock reason why there

dynamics. The bedrock reason why there is such an insatiable and growing demand for chips and other AI related hardware

is that the companies that are creating large language models all assume that they must keep scaling them bigger. The

bigger they get, the better they get.

Chap GPT2 is much better than chat GPT1 and so on and so on. And to achieve AGI, artificial general intelligence, the models need to scale ever larger and

larger. And the more scaling, the

larger. And the more scaling, the everinccreasing demand for chips and hardware. The first change in this armor

hardware. The first change in this armor occurred when OpenAI released Chat GPT5 in August, and critics pointed out that it was not that much better than Chat

GPT4.

Since then, there has been a growing but still minority opinion that believes that continuing to scale LLMs is

becoming a dead end. A few weeks ago, one of the original founders of Open AI announced that he had come around to this minority opinion. He believes that

continuing to scale LLMs has entered an era of diminishing returns and that the industry needs to go back to basic

research. Now, I'm the first to admit

research. Now, I'm the first to admit that I am no expert in computer science, but I can sense when a minority opinion is gaining traction. If this opinion

gains further traction, then some companies may back off scaling LLMs. Once that happens, the need for more chips will diminish and tech hardware

growth will slow across the board and the virtuous cycle will reverse. In my

view, the potential realization that the ever larger LLMs won't get us to the finish line is the fundamental risk to

the entire AI story. Theme three, the growth in private debt. It's a bull market and yet stocks like Apollo and Blackstone are down 10 and 12%

respectively. Why? Because investors are

respectively. Why? Because investors are worried about the growth in private credit and what potential losses lie underneath. The private credit debt

underneath. The private credit debt universe has ballooned from around 2 trillion in 2020 to roughly an incredible three trillion in 2025 with

US banks themselves lending close to 300 billion to private credit providers.

Now, there have already been two canaries in the coal mine, but they have not been plain old losses, but actual fraud. Triricolor was a used car lender

fraud. Triricolor was a used car lender that suddenly imploded. Creditors are

probing whether the same vehicles were pledged to multiple lenders. There are

reports of vehicles vanishing. First

Brands was an automotive aftermarket parts company that also suddenly evaporated. Lenders have made

evaporated. Lenders have made allegations similar toricolor, the reuse or misrepresentation of collateral. By

the way, the loans forricolor and first brands did not originate in the private credit world, but they are still viewed as warnings about what could happen.

Now, it's very hard to make a firm conclusion about the risks inherent in private credit. Before the great

private credit. Before the great financial crisis, it was actually fairly easy to see the deterioration in mortgage credit. In the United States,

mortgage credit. In the United States, mortgages are all securitized in public securizations. Every securization

securizations. Every securization reports its credit data every month and anyone can have access to that data if they are willing to pay Modi's or S&P

for access to their securization database. I did and I watched it happen

database. I did and I watched it happen here. Private credit is just that,

here. Private credit is just that, private. There are no databases that

private. There are no databases that show monthly or quarterly credit data on private credit. So, if it blows up, it's

private credit. So, if it blows up, it's almost by definition going to be a surprise. The market is afraid, but I

surprise. The market is afraid, but I don't think that the growth in private credit will cause a recession. What I do think is that when a recession

eventually does occur, the weaknesses in private credit, whether they are big or small, will be revealed. Theme four,

affordability. This has become a big political issue. Although inflation has

political issue. Although inflation has slowed, the cumulative impact of all that inflation over the last several years continues to be felt. And one can

see the impact of affordability in sectors like consumer staples and consumer discretionary which we will discuss soon. Let me just touch on the

discuss soon. Let me just touch on the affordability crisis in housing which I spoke about in a previous w. Sure, if

the 10-year yield goes below 4% that will help housing demand. Yet the

reasons as to why housing is so unaffordable go well beyond mortgage rates that are too high. My view is that this is not a demands problem. If we

could build affordable homes, people would buy them in a heartbeat. And the

usual solutions that politicians give are always related to the demand side.

During the presidential campaign, the Democrats proposed giving $25,000 in down payment assistance for first-time home buyers, but that would

not solve the problem because housing prices would just go up. This

administration is proposing a 50-year fixed rate mortgage. That won't solve the problem either. Housing prices would also just go up. And the other problem

with a 50-year fixed rate mortgage is that it will just take consumers that much longer to build equity in their

homes. Demand is not the problem. It's

homes. Demand is not the problem. It's

supply. And supply issues are almost all local issues. Now, what do I mean by

local issues. Now, what do I mean by that? It would be great if homebuilders

that? It would be great if homebuilders could just profitably build smaller homes, but they can't because many local

zoning laws have minimum square footage requirements that are just too high.

There are just so many local regulations to get land entitled, etc. that they add significant time and costs to building

new homes. Also, localities impose

new homes. Also, localities impose significant impact fees for new building communities. Now, those fees are

communities. Now, those fees are designed to defay local costs to support new home communities, but they can be

exorbitant and can add 10 to 15% to the cost of building new homes. This is why I know that politicians are not serious about the housing affordability problem.

Tackling local issues from a federal level can be a political mindfield. What

we need is a reduction in regulations at the local level, a reduction in impact fees, a reduction in the time for permit

approvals, and also regulations that will allow for the building of smaller single family homes. And the only way for the federal government to achieve

this is to threaten to cut off state and local funds. Unless they start moving in

local funds. Unless they start moving in that direction. Until you hear about

that direction. Until you hear about radical things like that from the federal government, the problem will not even begin to be dealt with. Anything

that does not attack the local supply problem cannot be taken seriously.

Hi, Steve Eisman here. On my weekly rap, I try to both teach and convey information as objectively as possible.

I try to make clear what are the facts and what are my opinions. But in today's media, it's increasingly hard to figure out what are the facts and where the

facts are being shaded by opinion.

That's why when I look into news events, I first go to ground news. Ground News

is my solution for getting to the facts of important stories, but also to see how left, right, and center are seeking to convey the same exact story. Take for

example the recent Senate rejection of dueling healthc care bills as the Obamacare deadline nears. To understand

the story, I went to groundnews.com and clicked on that particular story headline. There I immediately saw four

headline. There I immediately saw four tabs, left, center, right, and bias comparison. When I clicked on the center

comparison. When I clicked on the center tab, a series of headlines appeared, all from centerleaning sources. I could then click on any of those headlines and read

the story at the source. The same

happened when I clicked on the left tab and on the right tab. The bias

comparison tab showed Ground News's own analysis of how all three political leanings conveyed the same exact story.

I find the Ground News system enormously helpful because it allows me to easily separate the facts from opinions. I use

Ground News and I recommend you try it out. Go to groundnews.comal

out. Go to groundnews.comal

to get 40% off their unlimited access Vantage subscription. That's

Vantage subscription. That's groundnews.com/real.

groundnews.com/real.

groundnews.com/real.

And if you don't mind, use this link to get the discount so they know I sent you. Now, let's turn to how the 11

you. Now, let's turn to how the 11 sectors of the S&P 500 did this year and how each sector's performance reflects the four big themes I just discussed. On

the screen is a table showing all 11 sectors of the S&P 500 index. There are

three columns. The first column shows whether the particular sector is considered growth, cyclical, or staples.

The second column shows the waiting of the sector within the S&P 500. And the

last column shows the performance of each sector as of December 12th.

Generally, growth sectors did great infoch and communication services. Some

cyclical sectors did well because they contain companies related to AI construction. For example, industrials.

construction. For example, industrials.

Financials. A cyclical sector did well for reasons we will explore shortly.

Staples and defensive stocks did poorly for a host of reasons. They are not AI related. Also the consumer is showing

related. Also the consumer is showing stress and staples tend to be consumer related. The same applies to consumer

related. The same applies to consumer discretionary which also underperform.

Now let's dig deeper. Infoch a growth sector and the biggest sector at 35% of the S&P. The sector did very well up

the S&P. The sector did very well up 26%. But not as well as you might think

26%. But not as well as you might think given the dominance of the AI theme this year. That is because AI has already

year. That is because AI has already revealed winners but also some potential long-term losers thereby creating

tremendous dispersion within the info sector. So yes, Nvidia is up 30% this

sector. So yes, Nvidia is up 30% this year and Palunteer is up a whopping 140%. However, software and tech

140%. However, software and tech consulting perform poorly even though it has been a great space for decades. The

large software companies tend to have large and insurmountable modes around their businesses. Consistent and

their businesses. Consistent and modelable growth can be relied on. Also,

tech consulting has been a growth area for a long time. The current fear is that AI is reducing the cost of software production to such a degree that those

modes are no longer insurmountable and fear has gripped the sector. By the way, this is not a universally held view.

There are critics out there who say that AI does not really reduce the cost of software or not by much. But it's too early to really know and it's impossible

to disprove a negative. As for tech consulting, the thought is that such services are needed less when chat GPT can do much of the work. That is why

Salesforce is down 22% and Accenture and Gartner Group are down 23% and 52% respectively. Financials, a cyclical

respectively. Financials, a cyclical sector and the second largest sector at 14% of the S&P, which just goes to show how much tech dominates the United

States market and that the second largest sector, financials, is 60% smaller than infoch. Financials,

particularly large banks, did well because there was no recession. There

was the beginnings of an M&A boom and the yield curve steepened which helped the net interest margins of banks.

Cityroup was up 59% and Goldman was up 55%. The Lagards were alternative asset

55%. The Lagards were alternative asset managers like Apollo and Blackstone for the reasons already discussed. Payment

companies like Block and PayPal are down year-to date over 20% as the sector suffers from increased competition.

Communication services viewed as a growth sector and represents 10% of the S&P. It was the best performing sector

S&P. It was the best performing sector up 32%. Two areas outperformed. AI

up 32%. Two areas outperformed. AI

related stocks did exceptionally well.

Google was up 63% this year. Wow. Also,

traditional entertainment stocks also outperform as the auction for Warner Brothers up 180% lifted the entire traditional entertainment group.

Negative performers included cable and wireless companies as investors in feared increased competition and continued cord cutting. Consumer

discretionary 10% of the S&P 500, a cyclical group and a lagard up only 6%.

Why? Because it's a K-shaped economy.

And while the consumer continues to spend, there are plenty of weak spots emerging. In particular, housing and

emerging. In particular, housing and housing related stocks perform poorly with LAR, for example, down 12% this year. Also, while high-end consumers

year. Also, while high-end consumers continue to spend and Ralph Lauren is up 60% this year, the low-end consumer is suffering and pulling back. Hence,

Chipotle is down 40% this year.

Healthcare 9% of the S&P and up 11%.

Although healthcare underperformed the market, it is the best performing staples group. Some pharma stocks did

staples group. Some pharma stocks did well like Eli Liy up 33% which seems to have won the obesity drug wars. The

worst performing sub sector within health care was medical insurance as loss costs in all areas of health care

increased much more than expected. Thus,

United Healthcare, Sentine, and Molina were down 33%, 33% and 42% respectively. Industrials 8%

of the S&P and a cyclical group. The

group was up a nice 19%, but there was incredible dispersion within. Anything

AI related generally did well. Thus, the

best performing stock with industrials was GE Vernova. The GE spin-off. GEV

manufactures gas turbines for utilities.

Thus, its products are at the heart of supplying energy to AI database centers.

The stock is up over 100% this year.

Industrials related to housing performed poorly. Carrier Group, for example, an

poorly. Carrier Group, for example, an HVAC company, is down 22% this year.

Staples 4% of the S&P and up only 2%.

Obviously, Staples is considered a steady Eddie group and defensive stocks of all kinds have done poorly this year as investors chase the AI story. But the

poor performance of staples goes deeper than just that. Tariffs hurt margins and consumers are pulling back. That's why

Clorox, Campbell, and Kimberly Clark are down 36%, 32% and 28% respectively. The other sectors in the

respectively. The other sectors in the S&P are so small that they amount to little more than rounding errors.

Predictions are always difficult. So

where does that leave us for next year?

It's still a K-shaped economy whose growth is totally reliant on AI related capex. So far the market expresses its

capex. So far the market expresses its nervousness about AI by selling companies like Oracle. The market is worried about the debt Oracle is raising

to fund its AI capex and the debt markets are worried too. While Oracle is rated investment grade, its debt is now trading like junk. Now I think the AI

story will continue for at least another 6 months and then we will find out whether the two major AI risks I discussed will emerge. If the market

starts to digest these risk as real threats, we could see a major pullback in the overall stock market. AI has been the tide lifting many boats and the

reverse would not be pretty. Just like

in the financial crisis, the signs exist if you know where to look. Now, back

then, we had mortgage data from databases held by the rating agencies that we examine carefully. Today we have social media and that's where the tea

leaves can be found. Study discussions

on X and Reddit. When the tide turns, we will all say it was foretold on social media posts. If the tide turns, one huge

media posts. If the tide turns, one huge word of caution. Being too early on a call is not good. AI is still the most powerful engine the market has ever seen

in a very long time, and trends last longer than people predict. Calling the

end of a cycle of this magnitude is notoriously difficult for investors. The

answer is duration. If you have a very long-term duration and can ride out stock turbulence, then tune out and stay in. If your duration is short or medium

in. If your duration is short or medium turn, stay tuned in and stay nimble. You

don't need to be a hero. If you need the money that you have in stocks and can't afford to lose it, my advice to those investors is always have some percentage

available in cash. Federal government

money market funds return around 4%.

Cash doesn't have to stay in a bank account earning nothing. Cash can work for you without risk. And that's the year-end wrap. This last Monday, I

year-end wrap. This last Monday, I posted part one of my lecture on the causes of the great financial crisis.

And this Monday I will post part two.

Hope you tune in. This is the last wrap for the year. I want to thank all my viewers for your support in 2025 and for

your comments and participation. You

make the work exciting and challenging.

It's great to be in a conversation with many engaged and interested people.

Please, please keep the comments coming.

We will take a short break and our last episode of 2025 will be Monday, December 22nd when we post part two of my lecture

on the great financial crisis. Our first

episode for 2026 will be Monday, January 5th and the first wrap will be Friday, January 9. We want to wish everyone an

January 9. We want to wish everyone an amazing holiday season and happy new year. And I'm excited to share 2026 with

year. And I'm excited to share 2026 with my viewers and listeners. If you haven't already, please consider subscribing to our YouTube channel so you can receive

these weekly wraps along with our podcasts and the financial literacy master classes. Subscribing is the best

master classes. Subscribing is the best way to help the channel and we greatly greatly appreciate your support. Also,

be sure to check out our website realismanplaybook.com.

realismanplaybook.com.

There you can easily access all our episodes as well as the financial literacy master classes, our new blog, and some other goodies as well. Check it

out and see you soon.

This podcast is forformational purposes only and does not constitute investment advice. The hosts and guests may hold

advice. The hosts and guests may hold positions in stocks discussed. Opinions

expressed are their own and not recommendations. Please do your own due

recommendations. Please do your own due diligence and consult a licensed financial adviser before making any investment decisions.

Loading...

Loading video analysis...