HomeVideos

Nvidia is pointing to the next big AI trade

Now Playing

Nvidia is pointing to the next big AI trade

Transcript

249 segments

0:00

So, a lot of the talk around Nvidia has

0:02

been around circular financing. I think

0:04

it's also interesting to talk about it

0:06

in the context of what its portfolio is,

0:09

right? Because it is pretty unusual

0:13

for a large publicly traded company to

0:16

invest in other publicly traded

0:17

companies, including some of those that

0:19

are its clients. Um should people be

0:23

following the Nvidia portfolio and

0:25

investing in it for themselves?

0:27

>> Well, absolutely, Julie. I think what's

0:29

happening here is Nvidia is trying to

0:31

support the ecosystem and they're

0:33

investing in two ways. They're first

0:35

trying to support their suppliers. So,

0:37

anywhere where they see any sort of

0:40

bottleneck or something where they think

0:42

that this could be something that uh

0:44

will be really important to their

0:47

uh their build-out, uh they're investing

0:49

there. So, from that perspective, those

0:50

usually end up being very good

0:52

opportunities. And then on the other

0:54

side, they're trying to support

0:56

customers or people that are buying

0:58

their chips. And that's a little bit

1:00

more debatable, right? So, here is here

1:03

companies that maybe are not able to

1:05

secure financing on their own uh and

1:07

with Nvidia's support, they're going to

1:09

be able to get better credit terms. So,

1:11

I think if you're going to be following

1:12

what Nvidia is doing, I think the

1:14

telling tale is a little stronger uh on

1:16

the supplier side uh rather than the

1:19

customer side.

1:20

>> Um and you know, incidentally, at least

1:22

some of what Nvidia invests in is an

1:25

overlap with what you invest in um in

1:27

your in your ETF. And I've actually

1:29

created um recreated the top 10 holdings

1:32

in your portfolio, if I had them right,

1:33

on our Alphaspace platform here. A lot

1:36

of them are in these sort of optical

1:37

components companies, but you also have

1:39

the likes of Arm in there. You have

1:41

Coherent, which we mentioned. We have

1:42

Lumentum, which also uh just reported.

1:45

And also, if you take a look at the sort

1:47

of year-to-date performance for many of

1:49

these, you know, even if they've had

1:51

some more recent volatility, the

1:53

year-to-date performance, as we're

1:55

showing it here, has been pretty darn

1:57

good. The The white one here is

1:59

Coherent, which has been really an

2:00

outperformer

2:02

this year. So, how do you think about as

2:04

you said the suppliers are where you

2:05

want to be? Clearly, you're in the

2:06

suppliers. Um talk to me about that

2:09

thesis, which is you know, the sort of

2:11

like picks and shovels have been popular

2:13

for a while now. Where are we right now

2:15

in that cycle?

2:17

>> Well, Julia, what's been happening with

2:18

the picks and shovels is the bottlenecks

2:20

are shifting from some place to others.

2:23

So, it used to be going into this year,

2:25

memory was a big bottleneck. Going into

2:28

next year, we think optics is going to

2:29

have the same sort of dynamics is

2:32

equally cyclical to memory and there is

2:34

not enough supply. So, companies are

2:37

going to have to rush to add supply, but

2:38

they're not going to be able to add it

2:40

just in time. So, this is why optics is

2:42

one of the the trades we really like.

2:45

Capital equipment is another area. So,

2:47

as memory, as CPU capacity needs to get

2:50

added, you're going to see the semi cap

2:53

equipment names benefit. So, companies

2:55

like Lam Research,

2:57

uh KLA, those supply the equipment. So,

3:00

that's another area that this cycle is

3:02

just turning because companies are just

3:04

starting to invest rather than something

3:07

like the GPUs, they're we're already few

3:09

years into that that cycle.

3:11

>> So, okay. So, when you say they're

3:12

investing in this, so are we already

3:14

seeing these types of companies, whether

3:16

it's optical or the semi cap equipment,

3:18

are they already increasing capacity

3:20

with the anticipation that they're going

3:22

to see that I mean, they've already seen

3:24

a a bump in demand, but with the

3:25

anticipation that that's going to last?

3:27

>> Absolutely. So, for optics specifically,

3:30

most companies are planning to double

3:32

their capacity this year and then double

3:34

it again next year. But, what we're

3:36

seeing from demand, it could actually be

3:38

seven to 10 times bigger. And this is

3:41

because on top of their traditional

3:43

business, this used to be like telecom

3:45

suppliers with fiber optic cables, on

3:48

top of that, you have the data center

3:50

side, which has already been playing

3:52

out, but now optical components will

3:54

find their way into the servers

3:56

themselves for scale-up applications,

3:59

and it's an entirely new market that

4:01

didn't really exist before. So, this is

4:03

why the demand is growing 7x while the

4:06

supply is struggling to keep up with

4:08

even doubling every year. And yeah, this

4:11

is a trend we see across the board. So,

4:13

semicap equipment will be the companies

4:15

that provide the equipment for this

4:16

doubling of capacity.

4:18

>> Um obviously, you're well aware of the

4:20

sort of um

4:22

questions or concerns about this whole

4:24

cycle. There was an interesting um piece

4:26

this morning by a macro strategist at

4:28

Bloomberg named Simon White, who said um

4:31

he pointed specifically to the new

4:33

Nvidia financing structure, which it's

4:35

partnering with a bunch of Wall Street

4:36

giants to potentially put 500 uh billion

4:39

dollars out there in the system to try

4:40

and finance um more of this compute. As

4:43

he said in his piece, and he he was

4:45

drawing parallels between this and the

4:47

subprime housing crisis and AIG's

4:49

financing schemes in particular. And he

4:51

says, "In the housing crisis, increasing

4:53

financialization artificially boosted

4:55

demand, muting the signal from the

4:58

underlying market that there was already

4:59

too much supply in the aggregate. House

5:01

prices fell nationwide for the first

5:03

time since the 1930s, an event not

5:05

factored into any models."

5:08

So, do you have any concerns during this

5:10

cycle that because of the enormous money

5:13

being poured in the system, if there is

5:15

a change in demand

5:18

on the from the end user, that it will

5:20

be masked?

5:21

>> Well, Julie, absolutely. If there is

5:23

decline in demand when we have all the

5:25

supply coming in, these industries are

5:27

cyclical. So, this is why tracking the

5:30

demand side is so important. So, we talk

5:33

to companies either on the private side,

5:35

the public side, trying to assess how

5:37

much demand actually is there. And there

5:39

is also a way to track the spot pricing.

5:42

And right now, what we're seeing is

5:43

actually the opposite trend. So, people

5:45

are not necessarily chasing like with

5:48

housing, the difference is that people

5:50

are buying the doses and investments or

5:51

flips. Here, people are not really

5:54

flipping GPU capacity. They are buying

5:57

it for their internal needs. So, what

5:59

we're seeing with pricing

6:00

>> you could argue with most houses you're

6:02

buying it for to live in it, right?

6:04

>> Well, but when you see the big booms,

6:06

usually it's exacerbated by investment

6:09

type activity, right? So, it's not

6:11

necessarily like people didn't buy 10

6:13

houses to live in, right? It's really

6:15

going to be more driven. Like the

6:16

bubbles are more driven by investment

6:18

type activities, right? Same thing here,

6:22

if you start seeing companies that are

6:24

maybe like contracting GPU capacity and

6:27

they don't have the buyer, that would be

6:29

a concern. But as of right now, we see

6:31

more buyers for that capacity that need

6:34

these chips to to to use.

6:36

>> And what about the idea that the GPUs

6:38

that are already out there at data

6:40

centers are not at full capacity

6:42

utilization in many cases?

6:44

>> Well, that's something that is going to

6:46

change over time and that's basically

6:48

where we see a lot of opportunities here

6:50

today. So, companies that provide ways

6:52

for better GPU and CPU for that matter

6:55

utilization are really going to benefit

6:57

here. So, any company that So,

7:00

Cloudflare is one example where like

7:02

they have a a broad network of CPUs and

7:05

they help companies that want to build

7:07

AI applications use the CPUs

7:09

efficiently. So, these type of companies

7:11

are really going to benefit and I think

7:13

over time we will see utilization

7:16

increase significantly.

7:17

>> So, you talked about optical, you

7:18

stopped talked about semi-cap equipment.

7:21

Are there any other areas that you think

7:22

are I mean, it seems like

7:24

these areas are in general pretty well

7:27

understood, right? So, are there

7:29

anything Are there any other bottlenecks

7:30

that you see coming down the pipe that

7:32

would perhaps are not as anticipated

7:34

yet?

7:35

>> Uh well, I think the the optical

7:37

components is relatively newer and

7:40

semi-cap equipment is relatively newer.

7:42

CPUs as well, I think are not as well

7:45

understood.

7:46

We just that cycle just turned with AMD

7:49

and arm. This is why we own a arm. I

7:51

think as you see the agentic AI build

7:54

out, the AI agents will require a lot

7:57

more CPUs per GPU versus the data

7:59

centers themselves. So or model

8:02

training. So I think what we're going to

8:04

see is exponential increase in CPU

8:06

demand. So I would say from the data

8:08

center trade definitely CPUs, optics and

8:12

semi cup equipment are the places to be.

8:15

And then broadly speaking, we are very

8:17

excited about the space theme. I think

8:19

that's going to open a lot more

8:21

opportunities and new markets whether

8:23

it's defense, whether it's

8:25

communications or data centers in space

8:27

themselves. I think that's another

8:29

interesting idea to put on the radar.

8:31

>> Bonnie, good to see you. Thanks so much

8:32

for coming in.

8:34

And we also by the way just showed you

8:35

data from Yahoo Finance's Alpha Space

8:37

platform. That's a new professional

8:39

grade financial platform featuring

8:40

advanced charts, real-time news,

8:41

customizable investment research and

8:43

more. You can access all of those tools

8:45

by using the QR code on your screen.

Interactive Summary

The video features a discussion about Nvidia's investment strategy, focusing on their support for suppliers and customers in the tech ecosystem. The conversation highlights the investment opportunities in 'picks and shovels' companies, particularly in optical components, semiconductor capital equipment, and CPUs. The guest addresses concerns regarding potential market bubbles driven by excessive financing and emphasizes the importance of tracking end-user demand and capacity utilization.

Suggested questions

4 ready-made prompts