Hatem Dhiab on MU, SKHY & Samsung Bull Case, Memory's Lasting Bottleneck for AI

Nicole (0:00)

Hi, Hatem Dhiab's with me, managing partner, Gruber Kawasaki, Wealth Investment Management. Great to see you, thank you for being with us today.

So we had a lot of discussions on AI. Let's just start there, because we have the fears that we've now heard a lot about when we heard from Dario Emodi, we heard from Sam Altman and Elon Musk, Jensen Wong being on the other side of the coin. Some of your thoughts on the discussions we've heard in the last 48 hours or so.

Hatem Dhiab (0:29)

It's good to see you as well, Nicole. It's really wild. We have this year of pretty much the most successful new AI frontier company telling us to stop advancing.

I try to be level-headed when these conversations happen. I think the cat is out of the bag. We're not going to stop AI developments. I think what Dario is saying, which is Elon is saying and also Sam is saying, is we need some guardrails or some regulation. We need to understand how fast we need to go into the government to come in and really explain what's possible. Frankly, the development has been so fast that they all need to slow down a little bit.

Maybe it's self-serving because the spend on training has been so, so large that maybe slowing down all these frontier labs would be good for them, at least to the IPO. So I don't think it's going to stop the infrastructure build up that we've seen.

Nicole (1:34)

The issue too is that the infrastructure is behind the AI development, right? And that's how we come to pass that we need the infrastructure to go with what they're making. And so you don't want to slow it too much. This is what folks are saying. They don't want to slow it too much because how on earth would you have the infrastructure ready to go when needed?

At the same time, let's get to memory because that is the area that we want to talk about. We've talked about bottlenecks, shortages. What would you say is the theme on these memory chips this year? And what do you see going forward?

Hatem Dhiab (2:08)

Well, I think it's important to understand what's the span goes on. For every $1 spent on the data center, 50 cents will go towards your GPUs and all the ASICs.

20% goes towards memory.

10% roughly goes to all the networking, and 20% around construction and really everything else, land and so forth. That 20% going towards memory has gone up from 7% two years ago, simply because one, inference is huge. So, inference means more memory, more actually fast access memory, and then two, just prices have gone up so much because we don't have enough capacity. So, we're capacity constrained. So, I think that's why you've seen all these memory stocks have become really the hardest investment in the chip stack after the NVIDIAs of the world. I think that's going to continue for a few years because as you said, these investments, we're seeing 700, 800 billion a year. Jensen said almost 3 to 4 trillion over the next 3 years going towards data centers. I think about 20% of that going towards memory. So, the numbers are just huge.

Nicole (3:25)

In your notes, you talked about Korea's memory exports up 290%.

You said Samsung on top. Also, we had SK Hynix, another name, is a one to watch, as well as Micron, but the Korea story tells us that the demand is there, right? I mean, you have the demand, but now it's the bottlenecks and the shortages and how it bleeds into maybe disruptions?

Hatem Dhiab (3:53)

Well, I think, listen, Korea is really the number one player with Samsung and SK Hynix are the leaders there. So their exports are up 300 percent, but also their imports of memory production tools are up almost 200 percent. So they're building production because they know that demand is going to continue to be there for a long time. I think what's cool about the memory atmosphere right now is that before this used to be very, very cyclical business. You only could see three, six months out. Now we know what Micron is going to sell for the next three years. You know what's SK Hynix is going to sell for the next two or three years, right? All these companies have solid contracts at high, much, much, much higher prices with some of the best credit investors out there. So I think the reality is that demand is real and is real for many, many years to come.

So Korea being the bigger player there, and then you can obviously invest in all these companies through either ADR or the US with SK Hynix.