0:10

Hello, and welcome to another episode

0:13

of the Odd Lots Podcast. I'm Joe Wisenthal

0:16

and I'm Tracy Allaway. So,

0:19

Tracy, you know, we haven't talked

0:21

about it in a while, but one recurring

0:24

topic on the podcast over

0:26

the years has been China's

0:29

efforts to really sort of leap

0:31

into the lead technologically. We've talked

0:34

about it with a few different guests,

0:36

invested aggressively in airlines,

0:39

UH, semiconductors, UH

0:41

and so forth, setting us at

0:43

all the current noise. That's still like one of the bigger

0:46

long term storylines that attempt

0:48

to sort of supplant the US at the

0:50

lead in the lead of a technological

0:52

manufacturing and development. Yeah,

0:55

that's right, and I think that's where a

0:57

big portion of the trade

0:59

war, or the Trump administration's trade

1:01

war with China actually came from. And that's why

1:03

we see so much of it centered

1:06

around technology things

1:08

like Bye Dance, TikTok and UH

1:10

of course five G and Huawei and

1:13

UH. Of course we're seeing Beijing

1:16

rolling out it's it's new plans

1:18

for I think it's for the next twenty years or something

1:20

like that. And of course tech dominance features

1:23

quite highly. Once again, absolutely

1:26

but it also like it takes it takes

1:28

two to tango because in addition

1:31

to UH China's own endeavors

1:34

to leap frog or surpassed the US

1:36

in terms of technological development,

1:39

there's also an another element, which

1:41

is that a lot of the US companies

1:43

that we sort of associate with, you

1:46

know, the great heritage of US manufacturing,

1:49

whether it's Boeing, g E

1:52

or Intel with chips, all

1:54

of them seemed to be not doing

1:56

so great. So literally going in the other direction.

2:00

Yeah, so this is something that I wasn't

2:02

that aware of just because I've been outside

2:04

of the US for a while. But Intel

2:07

seems to be struggling. So I think

2:09

in the summer they said they were something like twelve

2:11

months behind schedule when it comes

2:13

to developing like their new next

2:15

generation of chips, and

2:18

the share prices come down quite a bit

2:20

since then. And it's sort of the polar

2:23

opposite to what you see happening at some other

2:25

chip companies like t S mc

2:27

over in Taiwan for instance, they seem to

2:29

be doing quite well during the pandemic, well

2:32

relatively well yeah, I mean you haven't

2:34

exactly right. There seems to be going in

2:36

the exact opposite direction, falling

2:39

further and further behind, and their

2:41

stock has really been got hammered this

2:43

summer. So, like what happened is

2:46

sort of an interesting question because we could talk about

2:48

China's research effort, but how

2:51

this company that basically invented

2:53

the industry, one of the you know, to put

2:56

the silicon and Silicon valley, so

2:58

to how it lost

3:00

its lead and ability to be one

3:02

of the world pre eminent manufacturers.

3:05

Is itself pretty interesting storyline

3:07

here. Yeah, absolutely,

3:09

And I mean I remember during the early two

3:11

thousand's Intel was this massive,

3:14

massive company and it had the branding

3:16

and the advertising around it and just

3:19

a huge deal. So it's interesting to see

3:21

it um sort of turning

3:23

around this year. So

3:26

we're going to talk more about Intel and what happened,

3:28

and I'm very excited about our guest.

3:31

We're going to be speaking with Stacy

3:33

Raskin. He's managing director

3:36

Senior analyst U s Semiconductors

3:38

at Bernstein Research, and

3:41

notably, at a recent uh

3:44

Intel conference call after one of their

3:46

quarters, the company didn't call on

3:48

him, leading him to a subtweet

3:50

a company on Twitter, which

3:52

is always fun and doesn't happen

3:54

that often with analysts. So Stacy, thank

3:56

you very much for joining us. Oh my pleasure.

3:59

I'm glad to be here. And by the way, um,

4:01

uh, nobody knows what that tweet was about.

4:04

Okay, it was

4:06

a subtweet of anything. Okay,

4:09

So just on the day that you didn't

4:11

get called on by until during

4:14

their quarterly conference call, you tweeted

4:16

cowards dot dot dot. But

4:18

we'll we'll all pretend that we have no

4:21

idea what that was about. Maybe it was something completely

4:23

unrelated to your professional career. This

4:25

is one of the mysteries of the universe. We'll

4:28

leave it at that. It'll be a mystery forever. We'll

4:31

just leave that. We'll just leave that tweet.

4:33

Um, but it is true that you didn't

4:35

get called on that part.

4:38

That is true. So let's forgetting

4:40

the subtweet aside or whether it's a subtweet,

4:43

what were you going to ask? Like, uh, you know, if

4:45

here's your chance, Stacy Raskin

4:48

at Bernstein Research Europe, what

4:50

would you have asked? Oh? Boy, I don't even remember

4:52

what I was going to ask at this point, and I have to go back and look

4:54

at my question list. Um, I'm sure

4:56

it was something pointed though, Um, it

4:58

typically as I have a

5:01

reputation I suppose for

5:03

asking those kinds of questions. And it's

5:05

not really my fault. I can't help it. My my

5:08

BS detection threshold is set kind

5:10

of low, and I tend to forget who

5:12

I'm talking to once I get going, So I can't help

5:14

it. But it doesn't always engender a

5:16

lot of love. Um, but that's okay.

5:19

I don't have to uh, I don't have to believe

5:21

this is gonna be a good conversation. Yeah,

5:23

this, this should be a great conversation. I'm looking forward

5:25

to it. Well, well, we'll see if that

5:28

threshold gets hit or not. I

5:31

feel like cell side analysts with low

5:34

BS detector is fairly unusual,

5:37

or at least we don't get to talk to many of them.

5:39

Um, it goes with the territory. Can

5:42

you set the scene for us, like, how

5:44

bad are things at Intel

5:47

at the moment? In your view, like Joe said,

5:49

you're sort of noted critic of the company

5:51

recently. What's going on there? How

5:53

bad is it? I mean, it's

5:56

it's problematic right now. And so you know,

5:58

you started out

6:00

the beginning of this call talking about

6:02

some of the issues they're having with manufacturing.

6:04

Um It's important to note

6:07

these are not new issues. So

6:09

the current issues are with seven nimeal

6:12

seven animeter process, and you're right. In July

6:15

they disclosed further delays there. They said

6:17

the process was was twelve months delayed. They

6:19

said the products, I think it versus

6:21

the prior road net were six months delayed. That

6:24

being said, this is not the first time they've had they've

6:26

had a problem ten nnimeters, which was the prior

6:29

generation manufacturing technology. And

6:31

if you want, we can go into what the numbers and everything

6:33

mean, but just in general, ten

6:35

anners was the prior one. They had big

6:37

problems with that one as well. That one was delayed for

6:40

five or six years and forced

6:42

them to sit on on the you know, the

6:44

two generations back, which was called fourteen animeters.

6:46

They were sitting on that one for many more years than they

6:49

intended. But even

6:51

fourteen nnimeters was delayed a little bit

6:53

when they when they originally launched it, it it was it took

6:55

about a year longer than they had originally anticipated

6:58

to reach sort of like fully manufacturable

7:02

yields and reach full volume. So these

7:04

are not new problems that Intel. These have been

7:06

building over probably half a decade

7:09

plus and they finally hit a wall.

7:12

So I mean that that's sort of the current state

7:14

of affairs. So Tracy and I

7:16

aren't technologists, but we see these terms

7:19

or hear them ten men, ten nanimator,

7:21

seven min nanimator. Why don't you take

7:23

this moment to describe like what these

7:26

means? I mean, I get it, okay, smaller and smaller

7:28

chips and efficiency and all that, but talk to us

7:30

a little bit about like what we're really talking

7:32

about. Yeah, But so in theory, like it

7:35

used to be, like some measure of the size of

7:37

the transistor there was called the gate length

7:39

that at one point um. In

7:41

reality, these numbers are complete in total

7:44

marketing. They don't actually mean anything in isolation

7:47

um, and they're not comparable from

7:49

one company to another. So, for example, an Intel

7:51

tenometer part is not the same as as a t

7:53

SMC ten nimet part. I'll

7:56

give you an example of this, by the way, And when when t

7:58

SMC went from what they purport really called

8:00

twenty nimes to sixteen, they

8:03

didn't shrink the transistors at all. What they did is

8:05

they changed the structure of those transistors.

8:07

Um. They went from what and again,

8:09

I we can talk about what these

8:11

terms mean. But they went from a transistor that was

8:13

called a planar transistor structure

8:16

to a three D or a fin fet transistor

8:18

structure. But they didn't make them any

8:21

smaller, They didn't squeeze them closer together,

8:23

they didn't do anything. They just went from planer to

8:25

fin fet and called it sixteen no

8:27

shrink, know nothing. So that's so that's that's the

8:30

thing that the members themselves actually

8:32

stopped really having actual

8:34

meaning or more than more than a decade

8:36

ago. It's been marketing ever ever since. What

8:38

you can think about them though, in terms of meaning,

8:41

it is typically some

8:43

measure of improved transistor

8:45

density. So squeezing more and more transistors

8:47

onto a given area of silicon, it is some measure

8:50

of that. And with every further note,

8:52

in every further transition in theory,

8:54

you should be getting better performance out of these

8:57

chips and and better power efficiency.

9:00

This is really what Moore's law is. By the way, it's sort of

9:02

these three legs in the stools. It's better performance,

9:04

better power, and lower cost. And by

9:06

the way, when people talk about Moore's law dying, which

9:08

is sort of the situation we're in right now. It

9:11

does not mean that it's impossible to to to

9:13

shrink these things to make the transistors any smaller.

9:15

What it actually means is that the cost leg of that three

9:18

legged stool is going out the window. So

9:20

we can still do this, like engineers are smart.

9:22

If there's a if there's a business case to

9:24

shrink, they'll do it. But now we have to pay for it, whereas

9:26

before it we used to get it every two years for free. It was

9:28

fantastic, right, But that's kind of the situation

9:31

of what's going on now. So

9:34

we keep comparing in this discussion

9:37

Intel to uh T SMC.

9:39

But my understanding is they're they're not exactly

9:43

the same company, right, the business model

9:45

is slightly different. Could you maybe explain

9:47

how they differ? You bet?

9:50

So. There are two primary um uh

9:52

semiconductor broad semic connector business

9:54

models in the industry. The first

9:56

is called they called an i DM or integrated

9:59

device man facturer, and these are companies of both

10:01

design and manufacture their own ships

10:03

like Intel. Right. The

10:06

issue is, you know, a leading

10:08

edge semiconductor manufacturing facility

10:10

can cost these days, you know, ten billion

10:12

dollars, Like it's very expensive. And

10:15

if you're going to build a factory that costs that much, you

10:17

better have a lot of revenue in order to put through

10:19

in order to cover those costs, and most most semi conduct

10:21

companies do not. And so decades

10:24

ago this became clear and the supply

10:26

chain atomized, it split

10:28

apart, and he gave rise to what's known as the fabulous

10:31

foundry model, where you have

10:33

companies, um say, like an Nvidio

10:35

or a qual Colm. These are referred to as fabulous companies

10:38

in the sense that they do not have fabs

10:40

fab as as a semiconductor manufacturing facility.

10:43

They only design chips and then they outsource

10:45

the manufacturing to a company known

10:47

as a foundry. This is what t SMC does,

10:50

and they're able because of this model. They can agglomerate

10:52

demand together from many semiconductor

10:55

companies and build up the revenue

10:57

scale that is required to support them

10:59

the high costs of of of manufacturing

11:02

um that ordinarily all these companies

11:04

would not be able to support on their own. And so that's what

11:06

t SMC does and it's it's been kind

11:08

of amazing, I mean in

11:10

the sense that you know, for the longest time,

11:13

Intel talked about being an i DM as being

11:15

an advantage because you could verily

11:17

tightly couple both the process and the design

11:19

together and get really good products.

11:22

And I guess as long as as both of those things

11:24

were on the right trajectory, that was true.

11:27

The problem is now because they're their their manufacturing

11:29

side has followed by the way side, it's actually really impacting

11:32

them. Not only can they not tightly couple

11:34

those two things together anymore, but

11:36

it interferes with the design because you designed

11:38

for a specific manufacturing process. If the process

11:41

isn't ready, you gotta throw that stuff out. And

11:43

so the fabuless founding

11:45

model, the founding model, it stuff actually seems to be growing.

11:48

It's ascending now. Um it seems

11:50

to have the advantage of especially because they are able to at

11:53

least at this moment, to stay on their manufacturing road

11:55

map. So we

11:57

can kind of think of Intel in

12:00

theory as you know, designing

12:03

chips all in video and

12:05

also having a fab t

12:08

SMC. What you're saying is it

12:10

sounds like you can't really like

12:13

sort of neatly divide the two

12:15

and if they're having trouble on the fab

12:17

side, on the foundry side, then

12:19

that also bleeds through to the design

12:22

side. Yeah, because you don't

12:24

design in isolation, right, you have to design

12:27

for a specific set There's a set

12:29

of design rules that go with a specific manufacturing

12:31

process. So, for example, if

12:33

I'm Intel, you know, and we'll probably get to this,

12:35

but until it's talking about using outsourcing

12:38

potentially in a bigger way, until

12:40

cannot just take their current designs and just

12:42

throw them over the wall to t SMC, they have to

12:44

completely redesign them um

12:47

to to correspond to t SMCS manufacturing

12:49

process, and and and and and t SMCS

12:51

design. So these two things are are coupled.

12:54

And if you have problems with one that causes problems with the

12:56

other. Sure, Wait,

12:59

could could we do of into the manufacturing

13:01

problems Intel a little bit more than like

13:03

what exactly is happening there and what's

13:06

gone wrong? Because of course, as we mentioned

13:08

in the intro, you know, Intel was supposed

13:11

to be the state of the art

13:13

global standard for chip making for

13:15

many many years, and now it seems

13:18

like it's not. Yes, So so first

13:20

I want to step back and and say, um,

13:23

this stuff is very very difficult

13:25

to do. So the fact that people are having problems

13:28

is not in and of itself a shock. These

13:30

are the most technologically

13:33

advanced products that humanity

13:35

has ever devised. And I don't know if either

13:38

of you have ever been in in a in a in a semiconductor

13:40

factory, but you can think

13:42

about this. I mean, if if if I just take these

13:44

nodes sizes as as gospel, which

13:46

I know they're not, but I mean, let's just just take them.

13:49

You know, at at Intel's right now

13:51

is delivering ten anometer products right when

13:53

when they're actually doing the manufacturing and they're

13:55

imprinting these these features onto

13:57

the onto the wafers. I'll try to be plaistic

14:00

here, but they use use a laser light to

14:02

to do this. The laser right now has

14:04

a wavelength of a hundred and nine three

14:06

nimes. So they're printing purportedly ten

14:09

nnimeter features using

14:11

a wavelength of light that's almost twenty times

14:13

the size of the feature that they're trying to print. Like,

14:15

I'm amazed that any of this stuff works at all. So

14:18

just just to get that out of the way, like it's it's

14:20

it's astonishing by the This is why one reason

14:23

that I love this space, because it's it's

14:25

like I'm just continually in awe of

14:27

the things that that that that humanity

14:29

collectively has has been able to pull together here.

14:32

But in terms of specifically what's going on it

14:34

Intel. So I'll step

14:36

back at fourteen nimes. Like

14:38

I said, that was a minor delay. Um

14:41

it was. Like I said, it took them about an

14:43

extra year to yield the process. Whether when I when

14:45

I say yields, that's a simple concept

14:48

that is, you know what, I'm making a silicon wafer

14:50

full of chips. How many of those chips are

14:52

good? That percentages your yield.

14:55

The more the higher yield, the lower your costs. So

14:57

that's a good thing, um. And it's the yield

14:59

that determine how quickly you ran for process

15:01

into production. Fourteen anometers took

15:03

a little bit longer. We still don't know why, but

15:06

it didn't hurt them because back then they had a genuine

15:08

five plus year process leadership. So they burned

15:10

a little bit of it. But it was fine. With

15:12

ten n we kind of know qualitatively

15:15

what was what was wrong, um until

15:17

had was using multiple patterning on

15:19

some of their layers for the first time. Again,

15:22

we can talk about what that is if if we

15:24

need to, but they were using a specific

15:26

type of of of advanced process

15:29

for the first time, and it caused problems. They

15:32

were using um new materials

15:35

at the time, like cobalt and other things, and

15:37

they were using other techniques

15:39

to get a much greater density improvement

15:42

that was normal with a normal node

15:44

transition. They their their transistor density

15:47

transitions permillimeter squared of of silicon

15:49

area typically went up by maybe two

15:51

point two to two point four times. With

15:53

ten anometers, it was a two point seven times

15:55

improvement. So the way the company has

15:58

discussed this is it was just too big of a lead and

16:01

they bit off more than they could chew and it caused problems.

16:03

UM, I have no idea what's

16:06

going on at seven animeters. All I know is that

16:08

it is it is something

16:10

that's completely different from anything they hit them at ten

16:13

and seven was supposed to fix those problems

16:15

attend it. It was a smaller density improvement,

16:17

it was only one point seven. They're

16:19

using a new lithography technique that's

16:21

called EUV or Extreme multra violet lithography,

16:24

that was going to replace that multi multi patterning

16:26

and fix those problems. And so they were

16:29

taking steps in order to to to

16:31

to learn lessons from the ten nnimeter debaucle

16:34

and and and to repair. And that that's why the seven

16:36

nimeter announcment came as

16:38

such a shock, because they've been telling

16:40

everybody, you know, for a year or seven animeters

16:43

is on track, we've learned the lessons, it's gonna be good,

16:45

and then they came out and just dropped the bomb on us. And

16:48

this is why going forward it's

16:50

so problematic, because their credibility on this stuff

16:53

is now zero. Right, so

16:55

when they're telling us where we think we've got to handle

16:57

on it, we think we know what's wrong. I mean, like nobody knows

16:59

what to believe anymore. So

17:16

many good questions. I think we're gonna have to extend

17:19

this episode to about two or three hours.

17:21

But in all seriousness, we

17:23

saw, you know, Tracy mentioned it when

17:25

they dropped this bomb this summer. Uh,

17:28

the stock took quite a reaction. I mean, it

17:30

was a major plunge. I don't think it's like a really

17:33

recovered very much since

17:35

then. Was that because this was

17:37

understood to be fundamentally different,

17:39

as you describe it, than those other delays,

17:42

Because there's this nobody really understand,

17:45

like explain the sort of financial market reaction

17:47

to this one versus the previous

17:49

setbacks. They've bet, yeah, you you bet, and

17:51

I need to full d A m D. I think and do this discussion

17:54

a little bit. Um. So if you go look

17:56

at A m D stock versus intels,

17:58

their polar opposites, right, and

18:01

and and part of the reason there was a A m D

18:03

is now on a trajectory and and they've come

18:05

a long way am D. You know, four five

18:08

years ago, the controversy on that stock was are they going

18:10

bankrupt or not? And the stock bottomed

18:12

out, you know, in below two dollars. That's

18:14

that's kind of where it was. The stock now was you know, close

18:16

to eight and it's been it's been even higher. The reason

18:19

is a A m D. Actually um uh,

18:21

they had already gone fabulous back in the financial

18:23

crisis. Um they had been using Global

18:26

Foundries, which is as another foundry

18:28

and Global Founderies back in the time

18:32

from shot am B in the foot um

18:34

Global Founders had problems with their process

18:37

technology as well. And then A m D. I

18:39

mean the business was in was in total free

18:42

fall. I mean they're there, their revenues a back cut in half.

18:44

They had four or five different rounds of restructuring was

18:46

pretty awful. They had a last Hail

18:48

Mary effort to get

18:50

a new architecture, which they're delivering now.

18:53

They're on their third generation UM and

18:55

that architecture was extremely successful,

18:58

and they made the shift from Global Founderies to t SMC.

19:00

And so now you're in a situation where where am

19:03

D actually has a competitive and

19:05

in some cases even superior architecture,

19:07

and they have a superior process technology. And so

19:09

they've been taking share from Intel both

19:12

on the client side pecs as well as in servers.

19:15

Right now, the real bulls on am

19:17

D. They have this vision right that

19:19

A and D is gonna go And I'll make up the numbers,

19:21

but I mean twenty

19:23

two there were twenty three. They'll be doing two

19:26

fifty or three bucks and earnings, and by five

19:28

they'll be doing five or six bucks and earnings. And

19:31

and this is a duopoli market. It's them an Intel

19:33

primarily, so all that would be coming at Intel's expense.

19:36

But am D has been kind of a dream.

19:39

And so before Intel announced these seven animeter

19:41

issues, it was kind of like, well, okay, maybe

19:44

am you'll get to this two or three dollar number. In

19:46

a few years. But by then a Intel will be going

19:49

harder and seven they'll have fix the issues. At ten

19:51

they gave am D an advantage. And at

19:54

that point, like it's all over right, and and

19:56

Intel is going to be back on on on the ascendancy

19:59

now because of these issues with seven

20:01

animeters, like with A and DNA, you you could believe

20:03

whatever you want, like there's there's no pushback

20:06

because you have no idea what Intel is gonna be doing at

20:08

that point um, and it makes the idea of

20:11

the the x C E six CPU market

20:14

and PCs and service, which today is sort of like a nine

20:16

ten or an eight twenty kind of market. You

20:19

know, if you wanted to believe that it could be

20:21

a true duopoly, a fifty fifth or sixty

20:23

market in fears, if you want to believe that, you can and

20:26

that all comes if it happens at Intel's

20:29

expense, and the driving force

20:31

behind it is a continued loss

20:33

potentially of competitiveness on Intel's part.

20:35

And with the seven animeter delay,

20:37

which like I said, was was really a bomb that was dropped

20:39

on people, it becomes very easy if

20:42

you want to believe that sort of thing, to believe it, So

20:44

this was potentially a thesis changing announcement

20:47

for Intel. That's where the stock reacted like it did. So

20:50

is this just I don't know? Is all of

20:53

this just down to Intel doing basically

20:55

a chip version of vertical

20:58

integration for far too

21:00

long when other companies

21:03

decided to specialize in various ways.

21:06

Yeah, so again, I don't know why Intel's

21:09

having this, but it's kind of amazing that they are, because they looked.

21:11

Intel is a hundred and thirteen thousand employees.

21:14

They spend over thirteen billion dollars

21:16

a year in revenue, which which is like almost

21:18

double amb's revenue. And in

21:20

R and D they Intel spends that much, it's almost double

21:22

ADS revenue. AD's growing. Now, I

21:24

don't, I don't know what's going on, like we we don't, We

21:26

don't know why. And now, by the way, it's it's causing

21:29

bigger problems because like Intel's face with a choice

21:31

now and and we'll find a little bit more about this in January.

21:34

Intel is now mate trying to make the decision. Are

21:36

they going to outsource it in in a greater fashion?

21:39

Right? Intel sources some don't, don't

21:41

get me wrong, Like in Intel's bought companies like

21:43

all Terror and Mobile I that use t smc

21:46

UM And for some of Intel's internal volumes,

21:48

you know, maybe not CPUs, but like chip sets

21:50

and peripherals and other things, they've they've used t SMC

21:52

So it's not new to Intel.

21:54

But I mean Intel is now talking about if

21:57

they can't fix their seven anime or problems, they

21:59

may have to scrap up it entirely, and

22:01

and outsources that that as well. UM

22:04

And so there's a lot of unanswered

22:06

questions with into right now. Like right now, I have no

22:08

idea what the company looks like in three years. That's

22:11

probably the biggest problem. Like if they were to come out

22:13

and give us a concrete plan and

22:16

give us certainty that they could execute on it, I think

22:18

the stock still would have fallen, you know, as much as it, but

22:20

maybe you could felt a little more comfort with buying

22:22

it at the bottom. Like right now

22:24

it is. It is very easily a valued

22:26

trap because again we we don't know what's going to happen in

22:28

three years. They're gonna make some announcement in January. We don't

22:31

know what they're gonna say. And in the meantime,

22:33

the competitive problems no matter what

22:35

they do, we're only gonna get worse. A and D is gonna keep

22:38

taking share on some trajectory. Apple

22:40

made an announcement yesterday, so like a good

22:42

chunk of I wouldn't be surprised if three quarters of

22:44

Apple's notebook business goes away from until next

22:46

year, which I'm not sure as in the numbers. I

22:48

know you didn't want to talk so much about short term, but

22:51

I still think numbers next year, I still think people

22:53

are smoking crack next year. The numbers are too high. They

22:55

need to come down. They may need to come down more. So

22:58

this is dangerous. You mentioned that you don't

23:00

know what until it is gonna look like in three

23:03

years. So how difficult does

23:05

that make your job as an analyst

23:07

in trying to determine, you know, an appropriate

23:10

price target or a way of valuing a

23:12

company where there is this much uncertainty. Well,

23:16

let me let me step back, because my job as an analyst,

23:18

like my favorite thing, is controversy, right,

23:20

because it gives me a reason to talk to my clients.

23:23

So um, from that standpoint,

23:25

this is fantastic, right, because

23:28

I don't think this controversy is gonna go away at any

23:30

time. In terms of sizing that I mean,

23:32

I mean, look like I have to sort of like our published

23:34

model right now is status

23:37

quo, and like I said, we're we're decently below

23:39

I think the streets too high. But one thing we can do

23:42

is sort of like run out scenario analysis and

23:44

like what could things look like? We We've done some of this, and

23:46

it's actually kind of interesting because you

23:49

know, if you sort of step back and you say, well, what would an Intel

23:51

look like if they were fablite or if they were

23:53

fabulous? If I had a magic

23:55

genie that just snapped her fingers and they were

23:58

fabulous tomorrow with

24:00

the same competitive environment as we have, it

24:02

doesn't necessarily look bad. Like gross margins

24:05

would be lower because they would be paying

24:07

a margin to to a TSOMEC to make the chips.

24:10

At the same time, they wouldn't be having to invest nearly

24:12

as much in their own factories, so you'd save cap as

24:15

you'd probably save R and D because you wouldn't be paying

24:17

for for research and development, for process

24:19

technology development, and so the operating

24:21

margins and free castles could be just as high as

24:23

they are and if not higher, Like that's possible if

24:26

you had a magic genie that made the transition

24:28

instant. The problem is, I don't have a magic

24:31

genie. It's gonna take years. Whatever

24:33

they do in the meantime, it's gonna throw more

24:35

uncertainty in the road map. And I'm not sure it's appropriate

24:37

to take today's competitive environment, which is already

24:40

deteriorating, and apply it

24:42

to that, you know, to as an overlay to that

24:44

model. And today where you have, like like I said,

24:46

a nine, ten or twenty model, maybe

24:48

by the time they're finished, it could very easily be a

24:50

sixty forty or fifty fifty model. And so

24:53

the more they talk about this, you know that it's

24:55

these are these are the kinds of things that you can roll out

24:58

as an analyst, and we've done that and we'll continue you to

25:00

do that. Let

25:02

me ask you broader question. I

25:04

mean, we we started talking about, we

25:07

started this conversation talking about the context

25:09

of the tech trade war, some

25:12

of the actions the current administration has made

25:14

against Huawei and so forth.

25:17

Should this be um, you know,

25:19

stepping aside from your your stock up

25:22

or down hat for a second or cell

25:24

side analysts had per se, but if

25:27

Intel a lot ends up

25:29

needing to rely on Taiwan semi

25:31

to make its chips. Is that

25:34

the type of issue that you

25:36

know could rise should rise to some

25:38

level of anxiety and DC, Yes,

25:41

I absolutely think it s because you think about

25:43

this right right now. You

25:45

know, broadly in the world, there are only a few

25:47

companies that can do leading edge semi manufacturing

25:50

anymore. It's it's Intel, Samsung

25:53

and TSMC. That is it. Everybody

25:55

else who has ever you know, been

25:58

at the forefront of semi connector manufacturing as

26:00

exitive the leading edge. That doesn't mean that other

26:02

players don't make stuff. I mean, there's there's plenty of folks

26:04

out there with fabs, but in terms of the bleeding

26:06

edge stuff, it's only those three companies anymore. Only

26:09

one of them is US and and the U S

26:11

one is having problems. And so if

26:13

you think about this, like what does that mean for Taiwan.

26:17

I mean they always used to say like data is the new

26:19

oil. I mean, maybe semi conductors are the new oil

26:21

now, and and Taiwan

26:24

like if this happens, I mean already is kind of turning

26:26

into potentially the most strategically important

26:29

country on the face of the earth. And

26:31

it's also a hundred and fifty miles offshore

26:33

from China and they kind of think they own the ground that it

26:35

sits on. So yeah, I

26:38

think this is potentially problematic.

26:40

And and and again, like even if Intel wanted to outsource

26:42

like the bulk of their leading edge to Taiwan, like, is

26:45

it politically viable? Like, I don't know, you

26:48

know, there's other things going on in the US now. I

26:50

mean, like one of the few sort of like bipartisan

26:53

initiatives that are out there right now is um

26:55

around strategic semi conductor investment.

26:58

Like certainly the Chinese are doing this. I mean, semis are

27:00

a big part of there have been a big part of their five year plans.

27:03

Um They've already been pushing towards self sufficiency.

27:06

Everything that's going on in terms of the trade

27:08

war and the sanctions and everything else is just going

27:10

to drive China to push towards

27:12

self sufficiency even more because we have them by

27:14

the balls right now, right, I mean,

27:16

if if, if if we wanted to completely

27:18

cut chinas semiconductor ambitions off of the

27:20

knees right now, we have the ability to do that by

27:23

banning semi sales. But it's a particular banning SEMIICAP

27:25

and and and and e d A and design software

27:28

sales. We're already kind of doing that to waway.

27:30

If we want to do it broader like, that's it. So

27:32

they have to move towards self sufficiency. And

27:34

I don't know what the political viability of of

27:37

having like the last bastion

27:39

of like advanced US based

27:41

semi eductor manufacturing being outsourced to a foreign

27:44

country. I I don't know how the how that's gonna fly, like, we'll

27:46

see, but it's got to certainly be on on top

27:48

of mind for any of the policymakers

27:50

who are looking at this area. On

27:54

that note, do you think there's a role

27:56

for policy to play in

27:58

I guess, helping out Intel if

28:01

if you think that it's a strategically important

28:03

industry for the United States, well

28:06

yes and no. So so there is have

28:08

always already been efforts to

28:10

invest and there were there was something called the Chips for

28:12

America Act, which we actually had bipartisan

28:14

support UM one of the few things

28:16

UM. I think that was subsumed into the National Defense

28:19

Authorization ACTUM. But there

28:21

is talk of of of funding for

28:23

semis now in terms of Intel's problems.

28:26

Like Intel's issue right now is not having enough

28:28

it is not not having enough money. They've got plenty

28:31

of money, So throwing money

28:33

at the problem is not going to fix

28:35

it, right, But you

28:37

know, could you make it easier in general

28:40

for them? I mean they're talking about things like um, you

28:42

know, investments for manufacturing, UM,

28:45

tax incentives, R and D incentives,

28:47

dark as getting involved that, so all of that stuff

28:49

would would be good. One issue I have with

28:52

some of the things that have been throwing around it is I still think they were dropping

28:54

the bucket. So we've seen a few

28:56

things. There's already some funny Now TSMC

28:58

is going to actually be building a tab in Arizona.

29:01

It's a PR headline right now. It's it's

29:03

like, I can't remember twenty thousand

29:05

wafers per month. Ultimately they're

29:08

gonna invest like twelve billion dollars or something

29:10

over eight years. I mean that that's a rounding error, frankly.

29:13

And then some of the numbers we saw in some of these initiatives,

29:15

there's really been no funding that's actually been um

29:17

apportioned. Yet they were talking about

29:20

like whatever it was twenty eight billion or thirty billion

29:22

or something that over a number of years. I personally

29:24

feel like, if the US is really serious about this, we

29:26

need an Apollo moment. We need like hundreds of

29:28

billions of dollars. I'm not gonna if we really

29:31

want to make sure that we we

29:33

we have you know, the majority of manufacturing

29:35

that it's rebased in the U S right now, Well, it's all moving

29:37

in Asia. UM. I don't know that

29:39

the political appetite

29:42

is there to invest that much money. UM,

29:44

but it's but I'll take whatever we can get at this point,

29:46

Like it's like whatever they want to do it to start. UM.

29:49

So I think, like long stories are think dollars

29:51

would would obviously would help Intel a number of

29:53

other players in the US just to make easy

29:56

but it won't fix that the technology problems

29:58

that are potentially driving the shift. Intel

30:00

us to do that on their own. MM.

30:17

One of the previous guests we've regularly

30:19

had on the show Dan

30:22

Wanang. He's a technology analyst at

30:24

gav Kell, and he talks about how like technoledge

30:28

comes in sort of two parts, or manufacturing knowledge

30:30

is in two parts. One is just sort of like blueprints,

30:32

which you can sort of like write down on a

30:35

piece of paper, and then the other is

30:37

just sort of like this like tacit sort

30:39

of understanding knowledge like how to

30:41

build a factory, which is not something

30:43

you just like write down on a piece of paper.

30:45

There's no like guy that someone can say, oh, this is how

30:48

you build a factory and just follow it. And it's something

30:50

that sort of like exists in the collective knowledge

30:52

of people who have been building factories

30:55

for their whole careers. So going

30:57

back to the seven animator to bacle,

31:00

is there something that's like identifiable

31:03

where like this some sort of like knowledge

31:06

or transmission of knowledge from one

31:08

generation of engineers is

31:11

getting degraded or lost within

31:13

Intel or within US manufacturing. Is

31:16

there some is there some sort of identifiable

31:18

leak where these things that once used to

31:20

be known and ongoing learnings and improvement

31:23

is just not happening to what it used to be. Well,

31:26

so I'll talk about like um, like

31:28

more general and more specifics. I mean, in general,

31:31

there's always a learning curve and so like like

31:34

I like, I'll make it up like I've got

31:36

ten animeters done and I'm going to seven, right,

31:39

seven an animeters is going to incorporate

31:41

all of the issues potentially that that that

31:43

the prior gen the ten had plus

31:45

a whole bunch of new ones. And

31:48

so the fact that they still haven't really fixed

31:50

ten and they and but they really haven't. They're

31:52

ramping ten animeters right now. But

31:55

it's it's it's it's impacting significantly

31:57

impacting their margins, which tells you that the yields

31:59

are not they have but they have to ramp it. They

32:01

can't wait any longer because of the competitive situation

32:03

where they and B right, so they've got

32:05

a rampant even though the yield aren't good. But if

32:08

you don't completely nail like the prior

32:11

note, all of those problems are gonna still pretty

32:13

much exist on the on the subsequent plus all to

32:15

new problems. And so that may be part

32:18

of it. And there's there's always learning curves on on on

32:20

this stuff, and if if you're learning curves too shallow, it

32:22

it can impact UM in terms of the employees.

32:25

And you asked like other issues with like knowledge

32:27

changers. So yeah, I I don't know specifically what's

32:29

going on at seven. I do know that

32:31

in intil had a very sizeable

32:33

layoff, and I

32:36

know that there were worries that they had been

32:38

laying off like some of the more senior

32:40

employees at the time, and there

32:42

were lawsuits and you can go back and look like there was a whole

32:44

bunch of stuff in the news flow around this. So I

32:47

have no idea what those employees were doing. I

32:49

don't know if they would have had anything to do with like the

32:51

subsequent seven enemies, I don't know, but

32:53

I do know that there was talk of a potential

32:55

brain grain several years back on

32:57

when when they had that lay up. So

33:00

who knows with with seven again, my biggest

33:03

concern with seven was again they never really

33:05

seemed to nail ten either, um

33:07

and, and that's going to to propagate and it seems

33:09

like it is. So we've

33:12

discussed all these problems, um and you

33:15

clearly have some concerns

33:17

about this company. How does management

33:20

actually go about restoring

33:23

confidence um in the business

33:25

model? Like, what what should they be doing at

33:27

this moment in time? Well, I

33:30

mean they're they're trying, right. I mean, look, there's there's

33:32

the magic solution, which is like I wake up

33:34

tomorrow morning and they say, hey, we fixed it. It's

33:36

all good. I guess that would solve things.

33:38

Although I'm not holding my breath or something like that. They're

33:41

doing other things. I mean, so you know they're

33:43

what they're doing now is they're they're trying to focus on other

33:45

areas for differentiation, and so it's there. It's

33:48

not necessarily just process, you know, it's

33:50

process. It's also things like packaging.

33:53

They're doing what's known as a chiplet architecture.

33:55

Now, where I could have a single product

33:57

that say we'll go into a PC someday, that

34:00

is not a single monolithic chip anymore. It's a whole

34:02

bunch of little chiplets that may be made in

34:05

different places, on different processes,

34:07

depending on whatever is best for that particular functionality,

34:11

and packaging those together in clever ways until

34:13

it's got good. I P there, until they're starting to leverage

34:15

that. They're also trying to leverage software

34:17

and other things. And like all of these things are all

34:19

important. They always were. Other players

34:21

are also good at some of those other things as well. But they're

34:24

they're gonna try the best they can, you

34:26

know, if process is not going to be the end all and be

34:28

all and everything to to try to leverage those those

34:30

other areas. And now, like I said, they're

34:32

gonna have to make a call like one way or the

34:34

other um in terms of what to do,

34:36

do you give up on decades

34:39

of of leading edge manufacturing? You know, you

34:41

know, prowess and and and leadership and

34:44

and bite the bullet and and and and and go

34:46

to TIS. I'm like, I don't know. I mean, they'll have to make

34:48

that call, and one way the other will will

34:50

know in a few years if it was the right call or not. Things are

34:52

gonna be pretty high, high profile decisions. I

34:55

also think tactically they need to reach personally.

34:57

And again, if we're just talking about the stock for a moment um,

35:00

I think they need to reset expectations. I think numbers

35:02

are too I and and the problem I have right now I mentioned

35:04

it once before. I don't know what it looks like in three years.

35:07

And not only am I below the street next year, but I

35:09

can't tell you with with with confidence

35:12

that next year represents the bottom either. And

35:14

that's the biggest problem. The stock itself is hard

35:16

to buy unless you feel

35:19

like expectations have been sufficiently reset

35:21

so they can grow off the trough. And I don't

35:23

think they're there yet, and I'm not sure so

35:26

far the current management team has not shown a willingness

35:29

to sort of like rip the band aid off, right, I mean

35:31

they've they've they've seems like they've wanted, you

35:33

know, to go in dribs and drabs, and that's usually not a

35:36

good way to go for stock performance, especially when

35:38

things are on the decline. So we'll

35:41

see what they do. There's the issue

35:44

just that you know, especially

35:46

as they don't you

35:48

know, they fall behind in UM, their

35:51

ability to YEA develop

35:53

cutting edge foundries potentially outsource

35:55

more. Is the issue just that Intel

35:58

will still be Intel, or Intel

36:00

will still be there and be competitive, but

36:02

it's just another chip company, and there's a Video

36:05

and t SMC and a m D and others,

36:07

and what was once the sort of category

36:09

defining company is just there's just another

36:12

player. Yeah, I mean, look, so I'm not worried

36:14

about until going to zero, like it's a behemoth

36:16

skill. I mean even today, go look at their numbers.

36:19

They're doing almost five bucks and earnings this year and delivering

36:21

I can't even remember the number, something like eighteen billion

36:23

plus in free cash flow. Right

36:26

and again, even even if it's on decline, like, those

36:28

are still pretty hefty numbers, right, And they

36:31

they've got a Ton, They've got a hundred and thirteen thousand

36:33

employees and they spend like listen, thirteen

36:35

billion plus a year in r D. It's hard to believe that they're

36:37

not they're not good things embedded within that

36:40

right that have value. But so

36:42

you have that. But but I think you're right. I mean,

36:44

if you're gonna go to a scenario where you know they're

36:46

no longer differentiated on processed, where

36:49

where their their destiny would at

36:51

that point would now be in t SMC's hands, you know, like

36:54

they're there their ability

36:56

to ship would be limited by how much capacity

36:58

they could get from t SM at that point, similar to how am

37:00

D is today. Right, that's a problem, um,

37:02

and where they're trying to differentiate and all these

37:05

things where other players are good as well, Yeah, you're you're

37:07

right, it becomes you know, they're they're

37:09

they no longer have that secret sauce necessarily

37:11

at that point they're competing you know, on a much more even

37:13

footing with lots of other players who are also very

37:16

very very good. And so you have that

37:18

that that kind of a problem. I mean, you can go look at

37:20

other tech comes like an IBM, for example, which

37:23

you and I be, and but I used to work at IBM

37:25

Research, like like way back in the days. I mean it's

37:27

it's it's a great company. They've got fantastic I

37:29

P and everything else. But I mean it's been a financial

37:31

engineering sort of thing for for many years. Right,

37:33

they haven't grown, you know, And

37:36

and and that's kind of what is they've they've they've been pulling

37:38

like increasingly financial levels and sort of manufacturing

37:41

earnings. And you know that that maybe

37:43

a scenario like if Intel can can kind

37:45

of get their mojo back, we may see something like that.

37:47

Um, it won't go to zero, but you're not gonna

37:50

see a ton of multiple expansion like with with with something

37:52

like that either, Right, And this is

37:54

the thing with Intel right now. I mean, like the bull case on

37:56

it from here is it is very cheap

37:58

and maybe something will go right. It's

38:01

cheap, hope, And both of those things are true. Like

38:03

it is extremely cheap even on my numbers

38:05

which are below it is inexpensive,

38:08

and maybe something will go right. I mean that that's

38:10

kind of ball, but that that is not like a hugely

38:13

compelling case. You never you never

38:15

sell a tech stock because it's just because it's

38:17

expensive. You never buy a tech stock just because

38:20

it's cheap. I think that's where we are with Intel

38:22

right now. Anything else,

38:24

sort of other last thoughts

38:26

Stacy that we didn't hit on that you think is

38:29

sort of important for understanding this story. No,

38:31

I think that covers it for Intel, Like if you if you'd

38:33

like to type talk broader and Semmi's I'm happy to

38:35

come back any time. By the way, well,

38:38

we'll have to do it again sometimes because I do feel

38:40

like that it's a fascinating subject

38:43

and I was pretty transfixed.

38:45

So I really appreciate you joining us, and

38:48

we'll have you bet you bet my pleasure anytime.

38:50

Yeah, thanks so much. Hopefully

38:53

hopefully you get to go on the next Intel call.

38:55

Huh. They kind of have to let you know.

38:57

Oh, I figure, like, look either either I'm

38:59

gonna be first in the queue where I'm never getting on again,

39:02

probably after that article. But

39:04

I had no comment to the reporter, like

39:07

I didn't comment on that, and I'm nothing in there was for

39:09

me. Well, thank

39:11

you for commenting to us. Yeah, yeah, I know You've

39:13

got any time, Tracy.

39:28

We got to do more of these sort of like

39:30

industry deep dives because

39:32

honestly like seven n animators,

39:35

ten nimators, all these issues, Like I feel

39:37

like, especially when you have someone like Stacy

39:39

who can explain these things, um, very

39:42

clearly, I could. I could really get hooked

39:44

up this stuff. Yeah, Stacy really stands

39:46

out. I think, Um, he sort of makes

39:48

self side analysis sound fun, doesn't

39:50

he. Like you can sense the

39:53

enthusiasm and what he does. Yeah.

39:55

Absolutely, And I think that, like you know, often,

39:59

um, maybe what sort of gets

40:01

lost uh in cell side is this sort

40:03

of like focus on numbers

40:06

and margins and all that stuff.

40:08

And I think that he does a really good job

40:11

of connecting the sort of like

40:13

the hard tech questions with

40:15

the margins. So thinking about like sort

40:18

of like Okay, they're ramping up ten nanometers

40:21

even though the processes aren't great, so the

40:23

yields are going to be lower, so the margins are

40:25

going to be lower. The connecting

40:27

the dots between the sort of technological

40:30

debt that they have, the technological difficulties

40:32

that they have with what actually falls through

40:35

to the bottom line. He does that really well. Yeah,

40:37

I agree, Um, and I really I mean,

40:40

having spoken to him, I now feel like

40:42

I need to listen to the next Intel

40:45

earnings fault just to see if they let him on or

40:47

not. Yeah, No, I'm interested that too.

40:49

Also like his point,

40:51

Um, you know, Taiwan is the most

40:53

strategically important place

40:55

in the world right now if like that's where all

40:58

of them you know,

41:00

that's where basically all of the cutting edge semiconductor

41:03

manufacturing is happening, or a huge

41:05

bulk of it. And it just feels like that

41:07

story is like, we gotta talk more

41:09

about it. I mean that's huge. Yeah, well

41:11

not only the technological angle, but also

41:13

the geopolitical angle as well. That's

41:16

sort of been heating up recently too. But

41:18

yeah, I agree with that. That's a good way of putting it. Um,

41:22

Okay, shall we leave it there, Let's save

41:24

it there. This has been another

41:26

episode of the All Thoughts podcast. I'm

41:29

Tracey Alloway. You can follow me on Twitter

41:31

at Tracy Alloway and

41:33

I'm Joe wisn't Thal. You can

41:35

follow me on Twitter at the Stalwart. Follow

41:38

our guests on Twitter. Stacy Raskin

41:40

he's at s Raskin. You can

41:42

see him sub tweet companies

41:45

or not. No one really knows exactly what he's

41:47

saying, but maybe he'll do it

41:49

again. Um so definitely

41:51

follow him there. Follow our producer Laura

41:53

Carlson. She's at Laura M. Carlson.

41:56

Follow the Bloomberg head of podcast brincesca

41:59

Levi at Prince just Get Today, and

42:01

check out all of our podcasts at Bloomberg

42:03

under the handle at podcasts. Thanks

42:06

for listening