TRANSCRIPTEnglish

DeepSeek "DeepSh*t" - Musk, Trump, and Panic.

40m 25s7,093 words995 segmentsEnglish

FULL TRANSCRIPT

0:00

should you buy the dip that happened in

0:02

yesterday's stock market and get back

0:04

into those chip stocks or add to those

0:06

chip stocks that just utterly tanked

0:09

yesterday what are the reactions of

0:12

people like Sam Alman Nvidia Elon Musk

0:15

Donald Trump what kind of new tariffs

0:17

are we now expecting including

0:19

potentially an Iron Dome and some other

0:21

executive orders will cover from Trump

0:22

we'll cover all of this related to deep

0:24

seek as well as what the popular

0:26

arguments are now remember yesterday we

0:28

had a reaction day and we're going to go

0:31

through today what the popular opinions

0:34

are now after people have had some time

0:36

to digest everything about deep seek so

0:39

we're going to look was this an

0:40

overreaction is this a buying

0:42

opportunity is this a real problem we'll

0:45

analyze it all we also I quickly will'll

0:48

get it out of the way I want to talk

0:50

about the micro strategy shelf uh and

0:52

some price action and then we'll also

0:54

have an update on the Tesla Optimus

0:57

towards the end of the video along with

0:58

our typical d Daily Wealth on today's me

1:01

Kevin report news that

1:05

makes uh so uh first micro strategy

1:07

yesterday uh they uh they issued an

1:10

update on their shelf offering it's

1:12

actually very interesting if you look at

1:14

it because what the Shelf offering shows

1:16

you is that they had over the period of

1:19

January 1 to January 26th spent about $

1:22

1.1 billion in cash to buy another

1:25

10,000 Bitcoin for a cost of about

1:28

$15.5 th000 per for Bitcoin uh and uh

1:32

they they had another 4.35 billion of

1:34

shares available for

1:36

issuance as of the morning of the 27th

1:40

in my opinion micro strategy actually

1:42

fell as much as 8% yesterday because

1:46

they sold stock to raise cash and then

1:49

they bought Bitcoin and as Bitcoin went

1:51

up because they hold you know over

1:52

470,000 Bitcoin as Bitcoin bottomed

1:56

around 2 p.m. so did micro strategy with

1:58

big buys coming in at 3 p.m. micro

2:01

strategy also following up I kind of

2:03

think it's interesting to look at some

2:05

people look at it and say oh my gosh how

2:06

much longer can they keep basically

2:08

selling Bitcoin for two times what it's

2:12

worth by selling their stock other

2:14

people buying it but then of course you

2:16

have the debate around oh but they have

2:17

software technology other people say oh

2:19

but they're losing money anyway it's an

2:21

interesting thing to pay attention to

2:23

but maybe not quite as interesting as

2:25

the potential that Elon Musk is pooping

2:28

bricks when it comes to deep seek yeah

2:33

we got to go into this one in detail and

2:35

look this is going to be entertaining

2:37

today's my birthday by the way so I'm

2:39

excited for the entertainment of a

2:41

birthday 33 we have uh meet Kevin coupon

2:45

code going for uh anybody who wants any

2:47

of the courses on building your wealth

2:48

trumponomics trade alerts lifetime

2:50

access to the course member live streams

2:52

check it out over at Meek kevin.com

2:53

we'll keep the pit short because it's

2:54

the birthday let's talk about Elon mus

2:56

pooping absolute breaks so xai raised

3:00

billion in May at a $24 billion

3:03

valuation uh and then they ended up

3:06

raising another 6 billion after that uh

3:10

I'm sorry another $5 billion at a $50

3:12

billion valuation in November so in

3:14

other words the valuation doubled and

3:15

they raised roughly the same amount of

3:17

money and they have about 100,000 h100s

3:21

that they use to train Gro the systems

3:24

called Colossus some say that Elon could

3:27

be really really quiet about deep seek

3:30

on purpose because he's freaking out

3:33

that the valuation at xai just got

3:36

reamed now I don't know that necessarily

3:39

this is true so what I wanted to do was

3:41

get some hints from what Elon Musk is

3:45

essentially up to on Twitter or x uh and

3:49

what I found was that Elon Musk didn't

3:52

actually tweet very much he at one point

3:56

expressed to Jade about uh potentially

4:00

uh this Chinese firm running deep seek

4:02

the hedge fund having a bunch of h100s

4:05

in other words suggesting that they

4:07

didn't train this on the older uh lesser

4:10

chips they trained them on h100s they

4:12

have a lot of and that's fine the

4:14

reality is whether they used h100s or

4:17

some older versions of chips that

4:19

doesn't actually matter as much for this

4:22

situation and you'll find that's true in

4:24

just a moment as we go through H that

4:28

that that may not have anything to do

4:30

with what's going on here but here were

4:32

some of his actual replies so look at

4:35

this and then help me understand what

4:37

you think he's up to so what do we have

4:40

here he retweeted jack dorsy saying open

4:43

source everything okay that's

4:46

interesting because xai doesn't open

4:48

source their uh Gro huh but retweets

4:52

that everything should be open sourced

4:54

okay then he trolls Jim Kramer on

4:56

nvidia's collapse because Jim Kramer

4:59

ended up tweeting on the 22nd that

5:00

nvidia's basically been trading sideways

5:02

and it's setting up for a breakout and

5:04

then Elon Musk yesterday wrote no

5:07

kidding you obviously trolling given

5:09

that in video was down 15 plus%

5:12

yesterday then the only person that Elon

5:16

Musk actually replied to about deep seek

5:19

was this guy Gavin Barker now Gavin

5:22

Barker in his message says that grock

5:25

looms large and he basically talks about

5:28

how amazing grock is and St tuned for

5:30

grock 3 because in a few weeks it'll be

5:32

epic and it'll be so good that it might

5:34

be able to cure cancer well that's a

5:37

very nice thing to say about elon's Gro

5:39

to which of course Elon replied this is

5:41

the only deep seek related like directly

5:44

related to deep seek reply I could find

5:46

Elon goes interesting analysis best I've

5:49

seen H okay so the best analysis you've

5:51

seen is the one that basically says

5:54

grock's going to cure cancer got it okay

5:56

and then AI spelled as AJ will be

5:58

everywhere except uh small little detail

6:01

Gavin Barker is also a reportedly large

6:05

investor in xai and a hedge fund manager

6:09

and he's been pumping up Gro for many

6:13

months oops it's kind of interesting so

6:16

this has a lot of people sort of putting

6:19

together this thesis that elon's been

6:22

pretty dead quiet on deep seek I mean

6:24

other than throwing Jade around what

6:26

chips they're using which really doesn't

6:27

matter making fun of Jim Kramer which

6:29

everybody does and then pumping up an

6:32

investor in xai he basically hasn't said

6:35

anything which is interesting because

6:38

most of the time you hear Elon Musk say

6:40

something like oh no no that's fake news

6:43

or that's a scam or something I actually

6:47

think Elon has looked at this model and

6:50

gone damn they actually did something

6:54

great with a lot

6:56

less are we wasting money on Nvidia

6:59

chips that's my take now that becomes

7:01

very very important for deep seek

7:04

because remember deep seek is all about

7:07

reducing the cost of either training or

7:11

even inference now that becomes

7:13

important that'll be a debate we'll talk

7:14

about in a moment but if companies led

7:17

by people like Elon Musk start going

7:19

dang we now have too many h100s or too

7:22

many chips we don't need this many chips

7:25

they're not going to dump these chips or

7:27

sell them they'll make do with what they

7:29

they have they'll figure something out

7:31

to do with them but they'll certainly be

7:32

thinking about how to become more

7:34

efficient in fact that's exactly what

7:36

Sam Alman said see Sam Alman actually

7:38

called the model impressive and then

7:41

went out of his way to say hey by the

7:43

way don't worry we've got some cool

7:45

announcements coming up including uh you

7:48

know basically more you know we'll be

7:50

working on making things more efficient

7:51

ourselves as well take a look at this

7:53

deep seek R1 is an impressive model it's

7:56

an open source model he could have

7:57

easily bagged on deep seek Sam Alman so

8:00

could Elon Musk but they're not the big

8:03

leaders in AI are telling you crap this

8:06

is this is actually really good deep

8:08

seek is an impressive model particularly

8:10

around what they're able to deliver for

8:11

the price we will obviously deliver much

8:14

better models and it's also invigorating

8:16

to have a new competitor so we'll pull

8:18

up some releases it's sort of implying

8:20

like don't worry we' we've got you know

8:22

things in the bag so to speak ready to

8:23

release but mostly we're excited to

8:25

continue to execute our research road

8:27

map and believe more compute is more

8:28

important now than ever before sure this

8:31

is one of the arguments we're going to

8:32

break down is more compute going to be

8:34

required now after deep seek than before

8:38

but fascinatingly you actually end up

8:40

also having Donald Trump then chime in

8:43

and say hey this is actually very uh

8:46

much a wake-up call to Americans we are

8:50

behind uh some VCS call this the Sputnik

8:53

moment where Russia was basically 10

8:55

years ahead of us in their lunar program

8:58

of course we did land on the Moon and

8:59

they didn't but anyway uh it was a

9:01

wakeup call and it was like wow okay

9:03

yeah other countries can actually get

9:05

ahead of us and so Donald Trump is like

9:07

hey this is a wakeup call we all should

9:10

figure out how to learn what we can here

9:13

and get better uh oh and uh by the way

9:16

Donald Trump ends up saying hey a lot of

9:18

companies will end up saving billions of

9:20

dollars this is great news so Trump

9:23

actually sees it from a typical sort of

9:25

like almost real estate developer where

9:27

he's like hey if you could save money

9:30

why spend more money like let's stupid

9:32

proof it for a moment if you have 500

9:35

nails and you get a permit or you only

9:38

need or want one house and you need 100

9:41

nails to build the house obviously I'm

9:42

oversimplifying here right are you going

9:44

to use more than 100

9:46

Nails maybe you'll use 110 because you

9:49

dropped a few right and obviously I'm

9:50

making up the numbers right but you'll

9:52

have about 400 extra Nails it's sort of

9:55

like having extra usage available to you

9:58

on chat GPT like are you really burning

10:01

through all of your AI usage that's

10:03

available right now let's say an

10:05

innovation comes along in construction

10:08

and it just takes five nails to build

10:09

your house together well I mean that's

10:12

wonderful you've just saved uh 95 nails

10:17

now you have an extra

10:19

495 nails and so then a lot of people

10:22

wonder like oh well that's great that

10:25

means people can build houses cheaper so

10:27

nail manufacturing will go up and I'm

10:30

like huh nail manufacturing AKA chip

10:34

development is going to go up because we

10:37

need fewer of them man this just sounds

10:39

Topsy Turvy and then I realize no the

10:44

only way nail manufacturing is going to

10:45

go up if all of a sudden there's a step

10:48

up in the Innovation that makes me want

10:49

now five times as many houses in other

10:52

words more artificial intelligence so

10:53

this is where we have to ask ourselves

10:55

is artificial intelligence going to

10:57

create somehow more demand

10:59

because all of a sudden it's better or

11:01

it's more Innovative or it could do more

11:03

for less if we could increase the demand

11:05

side then then yes we'll manufacture

11:08

more Nails we'll manufacture more chips

11:10

but if the demand stays the same so just

11:14

keep that very clear if demand stays

11:16

constant and cost goes down and we don't

11:21

need more demand because we're not

11:22

demand constrain there is no price

11:24

elasticity then then what ends up

11:27

happening is the people like on musk who

11:29

were buying the chips they just save

11:31

money like Donald Trump himself said

11:34

they save money so less people are

11:36

buying the chips fewer people are buying

11:38

the chips because they don't need as

11:40

many of the chips now this is all

11:43

incredibly important because again your

11:46

industry leaders aren't calling deep

11:48

seek a scam or a sop in fact even Nvidia

11:53

is applauding deep seek coming out with

11:55

a statement saying hey this is exactly

11:57

what our chips are designed to do

11:59

they're designed to bring efficiencies

12:01

into artificial intelligence so the

12:03

world can benefit hey they're taking the

12:05

high road in the day that their stock

12:07

was down

12:09

16.9% so no doubt here that deep seek is

12:12

real it's an open- Source model Elon

12:15

Musk isn't calling it a scam Nvidia

12:17

isn't calling it a scam Sam aldman isn't

12:19

calling it a scam uh you know Sergey o

12:22

over at um Microsoft isn't calling it a

12:25

scap so far everybody's arguing this is

12:27

impressive now what happens is the next

12:31

phase which is questionable the next

12:33

phase is okay who benefits where can we

12:38

go make money in the stock market as a

12:41

result of this and that's something that

12:43

we have to pay attention to because if

12:44

you look at it at least over this last

12:46

seven days take a peek what we have here

12:48

in Google search Trends you've actually

12:50

got deep sea outperforming chat GPT you

12:54

could really see that moment right here

12:56

where deep SE comes out and and takes

12:59

over GPT in just really the last day now

13:02

we'll watch this but let's understand

13:05

some of the arguments that go

13:08

into what's next for deep seek oh and

13:11

also quick note some people were asking

13:14

me about micro strategy and SD

13:16

STK that's the preferred offering for

13:19

certain people uh mostly institutions

13:22

and some retail I guess that they favor

13:24

or whatever that's not out yet that's

13:26

expected to price at the end of the

13:27

month anyway all right so

13:29

let's talk about the first big popular

13:31

argument so the one of the popular

13:33

arguments right now is chips are used

13:36

for inference not training and so all

13:40

people on the internet today are

13:41

freaking out because uh you know deep

13:44

seek makes it cheaper to train but wait

13:47

a minute you know we need chips for

13:50

training so don't worry if it's cheaper

13:52

to train we'll have more models demand

13:54

will go up and then we'll end up using

13:57

more chips because we want more

13:59

inference but this runs into a really

14:01

big problem there's the problem of

14:03

locality and this is actually where as I

14:06

mentioned on Friday as I mentioned on

14:09

Saturday and as I mentioned in

14:11

yesterday's Alpha report and the me

14:13

Kevin report apple is a style of a

14:16

winner here now this is where things get

14:18

really fascinating look at this chart

14:21

this chart shows you what it takes to

14:23

run different distillations of R1 on a

14:27

local computer now what's very important

14:30

to know is that the full-sized model is

14:33

671 billion parameters but one of deep

14:37

seeks Innovations is that you don't

14:39

actually have to run the full model for

14:43

inference in fact you only need to be

14:46

capable of running the 32 billion

14:49

parameter model and when you run a query

14:53

this is the innovation of R1 it selects

14:57

which 32 billion parameter segment of

15:00

the model to use so that way you're not

15:02

using the entire model you're just using

15:04

a portion of the model and it's the

15:05

portion that you need now I can run a 32

15:08

billion perimeter model with four Nvidia

15:11

490s I literally have those in this

15:13

house people are daisy chaining four or

15:16

five uh Mac minis together to run the 32

15:19

billion parameter model so the point

15:22

here is that deep seek actually has a

15:24

huge innovation in oh my gosh we don't

15:27

need to run the whole model we could

15:28

just parse out the section we need which

15:30

means compute required to run these

15:32

models is substantially lower than we've

15:34

ever thought before now we're not yet

15:37

where I could run the 32 billion

15:39

parameter model on my phone but I could

15:43

run the 8 billion parameter model on my

15:47

Mac I could daisy chain a couple

15:49

computers together and I could run the

15:52

14 billion parameter model and if and

15:55

and I could run a lighter version on my

15:57

phone soon enough though we will

16:00

probably have powerful enough phones and

16:03

we already have powerful enough

16:04

computers to run substantial versions of

16:06

these but soon enough my take is our

16:09

phones will be powerful enough to

16:10

locally run our own private version of

16:13

R1 and that I think is where people are

16:16

missing the boat people make the

16:18

argument in popular you know popularism

16:21

right now that the populist argument is

16:23

oh demand for NVIDIA chips is going to

16:25

go up because more people are going to

16:28

end up using AI I well first of all

16:29

nothing has changed on the demand curve

16:31

okay the innovation has not changed and

16:33

and I think that's what like drives me

16:35

nuts and I was talking about this

16:36

yesterday with course members is I was

16:38

with course members yesterday I drew

16:39

this like crappy drawing out I'm like

16:41

look AI used to be called neural Nets in

16:43

2018 then in 20123 we had the chat GPT

16:46

moment and all of a sudden you know we

16:48

got oh my gosh 3.5 and then we got four

16:51

uh you know 40 and this was great this

16:53

was a good Leap Forward in sort of chat

16:55

Bots which basically I call Universal

16:57

def def like I

16:59

drew a Define button here because I'm

17:01

like chat Bots are going to be as

17:03

ubiquitous and as much of a commodity as

17:05

a dictionary like do you really care if

17:07

you're using Mariam Webster or

17:10

dictionary.com 99.9% of people with the

17:13

exception of garans do not care what

17:15

dictionary like it's it doesn't just

17:17

give me a dictionary I just need to

17:18

define a word okay that's basically chat

17:20

Bots and the innovation has essentially

17:23

plateaued even the CEO of Microsoft has

17:26

said innovation plateaued he said it as

17:30

recently as December which is crazy

17:33

because I I I don't think that folks are

17:36

paying attention to this who say oh

17:37

demand is going to substantially

17:39

Skyrocket uh not only as the CEO of

17:41

Microsoft said it but I take took a a

17:44

screenshot of this here uh Ilia over

17:47

from open AI in November said that

17:50

pre-training has plateaued because

17:52

there's only one internet to train and

17:53

the largest llms already have been

17:55

trained on most of it now of course we

17:58

could use like Andre Andrew um Andre

18:02

kathri says we could use artificial

18:05

intelligence to create our own synthetic

18:07

data and then try to train off of that

18:10

but I don't know that we're there yet uh

18:12

that might be the next Innovation which

18:14

could be in 6 months 6 years or 30 years

18:17

but right now what's happening is we had

18:19

our GPT moment we've spent a lot of

18:21

money on chips and marketing deep seek

18:23

has come along to basically make this

18:26

spending look foolish and we still have

18:28

not had Innovation that's actually

18:30

increased demand that's the problem

18:33

that's the big thing that I think a lot

18:34

of people are missing in addition to

18:36

that fact that we could probably start

18:40

thinking about running these models

18:42

locally if we needed to so if we're a

18:44

small business we can now compete a

18:46

small business can very cheaply pick up

18:48

Mac minis or 490s or the soon to be

18:52

90s and run your model locally I

18:55

actually think it's a great business

18:56

idea you could set up a Consulting shop

18:59

and uh and Market hey here's your R1

19:03

local server for 15K you put the parts

19:07

together for 10 you take a 5k profit

19:09

preloading it all making it easy for

19:11

people you go sell it and Market it done

19:15

now people have local R1 and they don't

19:17

have uh privacy concerns by sending all

19:20

their data up to China or whatever model

19:23

takes uh these benefits of of scaling or

19:28

of how we're using portions of the AI to

19:32

be more efficient now this in part is

19:35

probably why apple is winning see apple

19:37

has pricing power because they have the

19:40

ecosystem you know they didn't blow

19:42

money on the llms no instead what they

19:46

did is they said hey we're just going to

19:49

keep trying to have and obviously

19:51

they're always behind but we're going to

19:53

try to have the best product that

19:54

eventually we could just run llms on

19:57

directly now the current AI they have

19:59

absolutely sucks there are memes going

20:01

around that Apple's AI is so bad that

20:04

their stock goes up when AI stocks

20:08

tank yes this is this is true right now

20:11

it definitely still sucks there's no

20:13

doubt about that but remember investing

20:15

is all about thinking about the future

20:17

and the future seems to be that running

20:20

AI locally as a commodity kind of like

20:22

you run a dictionary on your phone or

20:24

your iPad or your laptop is probably the

20:26

direction we're going and that is not a

20:28

benefit

20:29

like Nvidia is not a beneficiary of that

20:31

as much as I think people think I think

20:34

it's more the local companies the the

20:36

companies that can sell you a computer

20:38

basically they're potentially

20:40

beneficiaries now you have to be careful

20:41

with this because Apple sales like

20:43

iPhone sales in China are in a slow

20:45

cycle you're almost in a recessionary

20:47

cycle when it comes to PCS and

20:48

smartphones because sales are actually

20:50

declining so you kind of have to be

20:53

careful like oh my gosh okay I'll go all

20:54

in on Apple unless you see this as

20:56

temporary and buy the dip opportunity

20:58

the the thing is it's not just Apple

21:00

though it's zomi it's Huawei there are a

21:03

lot of competitors Samsung and they

21:06

could all be a part of that local device

21:09

provider so keep that in mind now again

21:12

uh people argue that demand is going to

21:16

go up this is this is the popular

21:18

argument I gave that nail example

21:21

because we don't really know if demand

21:23

is going to go up and even The Economist

21:25

reiterated this they said that deep

21:28

seeks Innovations suggest that The

21:30

Upfront cost of training a model may

21:32

plunge and now we're deploying more

21:35

computing power at the inference stage

21:37

when the model responds to questions the

21:40

models are able to now think or Reason

21:43

to give us better answers okay first of

21:45

all this whole reasoning thing is just

21:46

utter BS it's not reasoning what it's

21:49

doing is it's prompting for you so you

21:51

throw your question in and then it

21:54

breaks apart your question into little

21:56

prompts and it basically prompts itself

21:59

it's like an auto prompter remember in

22:01

2023 when everybody's like oh my gosh

22:02

you're going to have to become an a

22:04

prompt expert and prompt engineering is

22:06

going to be the biggest career in the

22:07

world or whatever I'm like that's bull

22:09

crap like you should not have to be a

22:12

prompt engineer and then sure enough the

22:14

reasoning models come out and they do

22:16

the prompting for you that's basically

22:18

it's not truly reasoning it's like faux

22:19

reasoning but anyway The Economist goes

22:22

on to say the the opposing forces of

22:25

Cheaper training but potentially priced

22:28

your inference have unclear impacts on

22:33

what's going to happen with compute

22:36

demand we don't know if more demand for

22:40

inference if we even have more demand

22:42

for inference is going to lead to more

22:43

chip spending if we could be so much

22:45

more efficient on the training side but

22:48

remember the economist doesn't mention

22:50

this R1 is so Innovative because it's a

22:52

cheaper to train but it's also cheaper

22:55

to infer from because again you're only

22:58

using a portion of the model you're not

23:00

running the whole

23:02

model again people are missing this

23:06

so my take uh is that we've got to be

23:10

very careful jumping to conclusions and

23:12

there's certainly going to be a lot more

23:14

time that goes into this but there are

23:17

other popular Arguments for example the

23:20

next popular argument is all of this is

23:22

just an

23:23

overreaction we still need chips and yes

23:27

like Bernstein wrote a big letter

23:28

they're like yes deep seek is great we

23:30

looked at the open source it's great

23:32

it's 20 to 40 times cheaper in other

23:35

words they admit that deep seek is

23:37

cheaper

23:39

and then they argue but we're still

23:42

going to need chips so please don't sell

23:44

Nvidia stock okay maybe that's true

23:49

maybe Nvidia will be fine but one

23:52

downside that I think people are

23:53

forgetting is that with the a without or

23:56

I should say with the absence of a new

23:58

innovation or a new Step Up in AI

24:01

technology what actually ends up

24:03

happening is you end up trending down

24:06

your Nvidia sales now I'm not talking

24:10

about a big plummet this is something

24:11

that takes time I think that if you were

24:15

expecting to have 30% EPS growth at

24:18

Nvidia over the next 5 years you might

24:21

actually see that slowly get written

24:23

down by analysts and then in reality

24:25

remember Nvidia sales were you know beat

24:27

estimates by two % uh last quarter you

24:30

might actually see Nvidia start missing

24:32

first they miss by 2% then they miss by

24:34

10% or then they guide lower and all of

24:37

a sudden EPS isn't growing at an average

24:39

of 30% it's growing at an average of 15%

24:42

then 10% then 5% and then oh interesting

24:45

it's starting to shrink it's not I I

24:47

don't believe that overnight people are

24:49

going to go cancel all the Nvidia orders

24:51

but I do think that this is a major

24:53

Catalyst that says we need fewer server

24:56

racks we need less energy we need fewer

25:00

chips and I think this is where if you

25:02

look at some of the other stocks that

25:04

got hit it's super micro computers got

25:07

hit well of course because if we don't

25:09

need as big a servers and we don't need

25:12

even more servers on top of what we

25:13

already have then of course super micro

25:15

is going to get hit of course Dell's

25:17

going to get hit of course arm Oracle

25:20

and Nvidia are going to get hit because

25:21

they're all beneficiaries of this

25:24

infrastructure spending continuing Sean

25:27

energy crashed like

25:29

20% other energy and uranium companies

25:32

VST ceg gev Olo they all dumped like 20%

25:37

yesterday due to the energy selloff

25:40

because markets are like okay I guess we

25:42

just don't need that much energy because

25:45

if it's local then you're running it

25:47

anyway probably don't need that much

25:49

energy because you're able to run it

25:51

locally and you don't need the high

25:54

performance computer the h100 but

25:56

honestly even if you did and went to

25:58

Blackwell remember Blackwell is 25% more

26:02

oh wait I'm sorry it's it uses 25% of

26:06

the energy so it's four times more

26:08

efficient than the

26:09

h100 so again you have this Duality

26:12

where the chips are using less energy

26:15

but now we potentially use less energy

26:16

to train but also less energy to compute

26:21

or to infer what the user wants why well

26:25

because you're only using a part of the

26:27

model that's the

26:29

Innovation now of course this is exactly

26:32

why people are saying okay well you know

26:36

uh then then sell All Tech sell all the

26:39

chips I don't know that you need to do

26:41

that you know I I never encourage sort

26:43

of panicking uh but it makes sense why

26:48

potentially the big growth estimates for

26:50

some of the chip plays could be over

26:52

another thing that to me kind of sends

26:53

the signal that you know deep seek is uh

26:57

is a big deal is yesterday DEC got hit

26:59

by a Cyber attack which has a lot of

27:01

people wondering ah did they get hit

27:03

with a Cyber attack because they're so

27:05

good that somebody's trying to take him

27:07

down I kind of think so now where else

27:10

are people going well Kathy Wood is

27:12

buying into Salesforce uh via arcg she

27:15

actually dumped Tesla to buy uh and

27:17

Palante actually to buy Salesforce

27:19

others are suggesting Adobe or docu sign

27:21

basically the software layer could be a

27:23

beneficiary uh I'm still not convinced

27:26

that the software products are that good

27:29

that uh we want them in Mass I mean

27:31

paler is pretty good but a lot of the

27:34

small business and medium business

27:36

related uh software problems or uh

27:39

products they run into the problem of

27:40

scale so they're really good at

27:43

individual tasks but when you try to

27:45

deploy them across an entire

27:47

organization they're still running into

27:49

scaling issues so it's going to be

27:51

something that we have to look for

27:52

companies to actually solve uh I think

27:55

paler has an edge there with the

27:57

government but the problem with paler is

27:59

people realize that finally uh I mean

28:01

I've been saying for two years that

28:03

Tesla and paler are like the real AI

28:06

that exists out there uh and markets

28:08

have finally adopted that thesis which

28:10

is great the stocks performed very very

28:12

well the downside is now the valuations

28:14

are very very very high now uh the

28:18

market ear they says we're they say that

28:20

overall we're in deep doodoo that we're

28:23

basically looking at

28:25

60% of the snps

28:29

29% return over the last 3 years coming

28:32

from the mag 7 and if it is true Kevin

28:35

that Nvidia earnings are going to go

28:36

down because we don't need as many nails

28:38

so to speak then remember Nvidia alone

28:42

out of those mag 7 was responsible for

28:44

25% of the return of the mag 7 and this

28:48

is where those bubble analogies come

28:49

from when very few concentrated stocks

28:52

pump up the indices what happens when

28:55

those earnings start faltering Goldman

28:58

Sachs talks about uh individuals going

29:01

into profit taking modes but potential

29:06

opportunities existing in Chinese stocks

29:09

since there're a lot with a lot of good

29:11

models moonshot AI 10cent bite Dance by

29:15

Mini Max

29:16

Alibaba many companies working Chinese

29:20

AI now I always question investing in

29:22

China mostly because I don't have the

29:24

insight into China that I feel like I do

29:25

in America but it's an interesting point

29:28

of view and a lot of people see uh some

29:30

potential in finding Chinese Investments

29:33

uh to make I personally think we'll

29:35

catch up in America and start using

29:38

segmented llms the way R1 does for

29:42

inference and also more efficient

29:44

training now uh the market year also

29:48

argues oh and Goldman says don't rush to

29:51

buy the dip both of them said that

29:53

Goldman argues that right now commodity

29:56

and trading Association stock Traders

29:57

are quote Max long and there's basically

30:00

no risk premium that you're earning if

30:02

you're investing in stocks right now so

30:04

they really question if investors start

30:08

becoming curious about artificial

30:10

intelligence capex and then leaders like

30:13

Elon Musk or Mark Zuckerberg or Serge

30:16

Bren over at Microsoft they start

30:17

pulling back spend and we're going to

30:21

have less uh we're going to have a

30:23

smaller set of earnings from the chip

30:25

related companies so be careful about

30:27

that uh Goldman says could it lead to

30:30

more spending in the long term maybe but

30:32

it could also lead to a big correction

30:35

first and when that sort of bleeds out

30:37

over time it doesn't have to be day

30:38

after day red you just kind of have to

30:40

wait to get the new estimates the

30:42

analyst updates and everything like if

30:44

Nvidia puts out a press release and says

30:46

hey by the way half of our orders just

30:47

got

30:49

cancelled markets are going to react

30:51

very quickly but it's not going to

30:52

happen today it's going to take weeks to

30:54

get this sort of

30:55

information uh so uh then we have uh

30:59

information uh we we talked about Apple

31:02

already briefly I just want to mention

31:04

that uh The Economist points out as well

31:08

that volumes and semiconductors outside

31:10

of AI PCS and smartphones still

31:13

suffering so just also keep that in mind

31:16

if you're investing in some of those

31:17

aspects the PCS and smartphone sectors

31:19

they're just not doing great right now

31:21

they still in a bit of a recessionary

31:23

environment uh UBS argues that uh

31:28

ultimately we could see innovations that

31:31

end up driving more demand for chips but

31:34

again I like to argue well show me the

31:37

Innovation because so far I've been

31:38

looking at AI Bots for the last couple

31:41

years and I feel like after 40 we didn't

31:44

get that far ahead but then again maybe

31:46

you have a different opinion I mean yeah

31:47

look we're able to like search the

31:49

internet now originally we weren't able

31:50

to do that we had dat information but

31:54

that was sort of an obvious step forward

31:57

now uh Standard Charter comes out with

32:00

uh sort of moving on if you will from uh

32:02

deep seek for a moment although what is

32:05

also kind of remarkable is deep seek did

32:07

end up coming out uh with a uh Vision

32:10

deployment model to uh compete with uh

32:13

Sora and uh mid journey and some of the

32:16

other uh image generators and people are

32:19

like how do they possibly keep coming

32:22

out with more products when they're

32:24

already dominating like this of course

32:26

that is leading to some meme circuit

32:28

ating like a picture of Sam Alman on the

32:30

phone going launch the porn Sora

32:34

anything to basically get it up I mean

32:36

get demand up but anyway let's move on a

32:40

little bit from this let's talk about

32:41

the fomc so we've got the fomc meeting

32:44

tomorrow uh there's a small indic well a

32:46

standard Charter argues that any

32:48

indication that we potentially get a

32:50

interest rate hike would end up being

32:52

very bad uh they argue that any kind of

32:55

firming in the labor market could end

32:57

end up reiterating the potential for

33:01

hikes and equities would sell off under

33:04

any indication of interest rate hikes

33:07

especially after this sort of deep- seek

33:09

moment any kind of optimism on

33:11

disinflation though would be very very

33:13

helpful to to markets so prepare for the

33:16

FED tomorrow and this is sort of a

33:18

little warning on the FED I guess uh

33:20

personally I don't think there's much of

33:21

a risk that Drome pow's going to talk

33:22

about hikes especially after the Deep

33:24

seek moment I think it's much more

33:26

likely that they're going to actually

33:28

applaud the disinflation that comes from

33:30

a deep seek moment because that's like

33:32

what Trump says hey you can all save

33:34

billions of dollars now after all the

33:36

anthropic CEO says it cost about a

33:38

billion dollars to train a model these

33:40

days if deep seek did it for 5.6 that's

33:43

great it's like pure deflation this is

33:46

actually wonderful for Jerome Powell

33:48

come on baby turn the money printer back

33:49

on we need

33:51

it then you got Donald Trump he

33:53

announced four new executive orders an

33:55

Iron Dome to protect America trans

33:57

genderism out of the military stopped

34:00

the indoctrination of our uh troops with

34:03

critical race Theory and offered full

34:06

reinstatement for those expelled due to

34:09

co uh you know refusing to get a covid

34:11

vaccine and people would be reinstated

34:13

with their former Rank and full pay he

34:17

also somewhat suggested by referring to

34:20

1887 that we used to have so much money

34:22

from collecting it from tariffs that we

34:24

were able to get by without an income

34:26

tax some people then sort of took that

34:28

and said oh my gosh he wants to get rid

34:29

of the income tax he suggested that in

34:31

the past but I think we got a long way

34:33

to go let's see how we do with some

34:35

tariffs first bassent was approved and

34:38

some people think that he's actually in

34:39

favor of some form of universal tariffs

34:42

and some say we could see tariffs as

34:44

soon as this weekend on Mexico Canada

34:47

maybe even China you also by the way

34:50

have Donald Trump suggesting placing

34:52

tariffs on chip manufacturers outside of

34:54

the US this would actually include

34:55

Taiwan and he argues if you don't want

34:58

tariffs build your plants in the United

35:00

States Morgan Stanley had a bit of a

35:03

piece on Optimus and they're actually

35:05

this is Adam Jonas who used to be a

35:07

Tesla bull he actually doesn't include

35:10

Optimus robots in his 2040 valuation

35:12

model because he says the Battery

35:14

Technology just isn't there that current

35:17

humanoids really only run on about 50

35:19

minutes maybe up to 2 hours of battery

35:22

time and unless you want them in a

35:24

factory plugged in all the time you're

35:26

not really getting functional humanoid

35:29

robot uh and uh therefore it's going to

35:31

be a while before we actually get these

35:34

functional now uh a lot of people think

35:37

uh and listen to rumors that Elon Musk

35:39

wants to manufacture as many as 600

35:41

Optimus robots by the end of the year uh

35:43

per week it could be possible and in

35:46

line with his goal to have a lot of

35:47

Optimus robots working in uh Tesla

35:50

factories and that might be the perfect

35:52

place for them because they could stay

35:53

plugged in but in terms of a mass Market

35:56

it does raise the question question of

35:58

when are we going to be willing to have

36:00

a robot that has to charge itself every

36:03

hour hey if it could charge itself maybe

36:05

it's not that bad just go sit on that

36:07

chair over there there's a there's a

36:08

pointy thing just sit on that pointy

36:10

thing there you'll be fine other people

36:13

uh pay a little bit of attention to

36:15

Tesla now suing the European Union about

36:18

tariffs uh Tesla uh does get tariffed a

36:22

little bit less than some of the other

36:23

Chinese makers in the European Union and

36:26

Tesla counts for 28 % of all Chinese

36:28

made uh EV Imports into the European

36:31

Union but I don't know that this lawsuit

36:34

is really going to go anywhere in my

36:35

opinion this is messaging I don't

36:37

actually think I'm not very optimistic

36:39

that they're going to succeed in this

36:41

since it is their right to implement

36:43

tariffs and Tesla has lower tariffs than

36:45

the

36:46

others but Tesla's still trying it's

36:49

probably a messaging thing uh to sort of

36:52

uh you know be Pro Tesla and uh and

36:54

anti- the EU which seems to be a pop po

36:57

argument by mus these days uh anyway the

37:00

new model y does have about 3 to 7% more

37:03

range uh front camera for better parking

37:05

and some better Lighting on the exterior

37:07

some cool light bars we are looking at a

37:10

potential consensus decline in margins

37:14

for Tesla earnings coming up looking for

37:17

about 15% on uh gross margins in Q4 that

37:21

would be down from about 15.6% in Q3

37:24

excluding those FSD deferred revenues

37:26

which those did bump their margins up

37:29

but you kind of have to parse that out

37:30

which doesn't happen on earnings day you

37:32

have to do that a couple days later when

37:33

they actually release uh their 10K or

37:36

their 10 Q anyway big things we're going

37:39

to be looking for are going to be

37:40

progress on Robo taxi development

37:42

Optimus progress battery progress

37:44

especially in the face of this uh and uh

37:47

we'll keep an eye on uh what ends up

37:49

happening we'll be covering earnings of

37:50

course the vake kind of hinted now more

37:53

officially on Jesse wat show that uh he

37:55

left Doge on good terms he's going to

37:58

divide and conquer because he wants to

37:59

run for elected office uh then we've got

38:03

to hit the dad joke of the day so Dad

38:05

joke of the day I'll start with the the

38:06

the joke portion I'm going to hit the

38:07

Daily Wealth and then we'll do the punch

38:09

line so get ready for it I think my

38:12

wife's been putting glue all over my

38:15

firearms keep that in mind let me

38:17

quickly hit the Daily Wealth Daily

38:19

Wealth for today is that there's a story

38:22

of a teacher locally here in Ventura

38:24

California who used to ask his students

38:26

every day uh sorry not every day every

38:29

year when they went to graduate

38:30

essentially to the next uh uh grade this

38:33

was about 10th grade so the students

38:34

were 15 16 years old hey what do you

38:37

want to be when you grow up as as you're

38:38

leaving and graduating from this and he

38:41

used to get answers like doctor

38:43

astronaut firefighter teacher fireman uh

38:48

software engineer mechanical engineer

38:50

pilot whatever used to get these sort of

38:53

careers that were respectable and uh and

38:57

and people look forward

38:59

to recently he said uh you know this

39:03

this has been for for decades he's been

39:04

getting answers like that so nothing

39:06

nothing new and

39:07

surprising more recently however he's

39:10

been getting answers from younger

39:11

Generations like I want to be a

39:14

billionaire and it's over and over and

39:17

over again and some people are looking

39:19

at that and wondering is social media

39:21

making us more relative than ever before

39:24

where we're constantly comparing

39:26

ourselves to others to where we're

39:28

basically never satisfied with what we

39:31

have kind of sad and it's an interesting

39:34

comparison and it really should take

39:36

give us an opportunity to look at what

39:38

we have and say we are we actually

39:40

grateful for what we have are we

39:42

grateful for the position that we're in

39:44

and what are we doing to improve our

39:47

position because we're always thinking

39:48

about Improvement uh but have we gotten

39:51

to the point of never being satisfied

39:54

and are we therefore making ourselves

39:56

miserable

39:58

interesting so now for the dad joke uh I

40:00

think my wife's been putting glue all

40:02

over my firearms she denies it but I'm

40:05

sticking to my guns do not advertise

40:08

these things that you told us here I

40:09

feel like nobody else knows about this

40:11

we'll we'll try a little advertising and

40:12

see how it goes congratulations man you

40:14

have done so much people love you people

40:15

look up to you Kevin PA there financial

40:18

analyst and YouTuber meet Kevin always

40:20

great to get your take

UNLOCK MORE

Sign up free to access premium features

INTERACTIVE VIEWER

Watch the video with synced subtitles, adjustable overlay, and full playback control.

SIGN UP FREE TO UNLOCK

AI SUMMARY

Get an instant AI-generated summary of the video content, key points, and takeaways.

SIGN UP FREE TO UNLOCK

TRANSLATE

Translate the transcript to 100+ languages with one click. Download in any format.

SIGN UP FREE TO UNLOCK

MIND MAP

Visualize the transcript as an interactive mind map. Understand structure at a glance.

SIGN UP FREE TO UNLOCK

CHAT WITH TRANSCRIPT

Ask questions about the video content. Get answers powered by AI directly from the transcript.

SIGN UP FREE TO UNLOCK

GET MORE FROM YOUR TRANSCRIPTS

Sign up for free and unlock interactive viewer, AI summaries, translations, mind maps, and more. No credit card required.