xETFs on Adding Portfolio Exposure to Samsung, SK Hynix & Memory 'Picks and Shovels'

SPEAKER_1 (0:00)

So, let's bring in our next guest now, Kenneth Wong, co-founder, CIO, XETF. Good afternoon, Kenneth. Thank you so much for your time. So, we do talk to you folks quite often over at XETF. So I've just noticed that in the last couple of days, you've introduced a new bottleneck ETF. It's called NEC, N-E-C-K.

I'm just wondering exactly where you see the biggest constraints right now.

Kenneth Wong (0:26)

Yeah, absolutely.

I think for us, we see many different constraints. Particularly, there are five that we are focused on within this NEC ETF.

The first is memory. I think before commercial break, you guys talked a little bit about kind of the Korea trade. We think Korea really is at the heart of it. And we are seeing a huge demand for memory. Most of the supply is sold out through the next year. And I think even SK Hynix's CEO said that, they see demand exceeding supply through 2030 Other areas that we are focused on would be optics and photonics. We're also focused on power. There's a lot of energy demand. All these data centers require a ton of power in order to keep them running for training and for inference. And then lastly, also from semiconductors and compute. So obviously NVIDIA is the big trade there, but besides kind of GPUs, we also see demand for kind of more custom GPUs. Also in terms of pure compute from like a Neo Cloud perspective.

SPEAKER_3 (1:32)

And you also have your KSMH ETF with the concept that Korea is also a central part of this AI memory build out. What makes the Korean semiconductor ecosystem particularly important to the broader AI infrastructure story?

Kenneth Wong (1:52)

Yeah, so Korea really is at the heart of the memory trade.

So about 80% of high bandwidth memory is currently being produced out of Korea right now, led by SK Hynix and Samsung. And the interesting thing about this is that it really isn't just a two company story, right? As you mentioned, there really is a broader ecosystem behind it. And that's really where KSMH comes in, where it's trying to access all the picks and shovels behind the two biggest companies out there. And there's a pretty good parallel that we can make to the US as well, where you see kind of the large hyperscalers and frontier AI labs, right? But behind them, we actually see a whole ton of beneficiaries in the infrastructure build out, whether it's kind of the NVIDIA, a lot of the bottleneck companies that I mentioned earlier before. That parallel, you can draw to Korea as well, where behind SK Hynix and behind Samsung, you have a whole bunch of other companies supplying the materials, the testing, the substrates, the circuit boards. All of that works in one big ecosystem in order to have this huge memory trade that is going on right now.

SPEAKER_1 (3:12)

And obviously, that memory trade has proven how this sector, I suppose, is a subsector, is moving away from the traditionally cyclical kind of nature that has always been characterized by. I'm just wondering though, the constraints can't last forever, Kenneth. So how do you think about them when that position isn't really permanent and things move on?

Kenneth Wong (3:40)

Yeah, absolutely. I mean, if you take a look historically, memory has generally been a fairly cyclical trade, but we are starting to see it move away from that. I think firstly, the scale of demand that we're seeing now is completely different from what we've seen before in the past. We've had memory up cycles in the past, but never to this extent. So for example, just this year, I think the memory market is supposed to quadruple.

We've seen DRAM prices go up five to seven times. We've never seen that in the past. As I mentioned earlier, everyone is expecting demand to outstrip supply through next year, really through 2030 as well. I think a second part of it is that a lot of these companies are starting to enter into more longer-term agreements. So rather than just trading memory chips at the spot price, they're actually locking in agreements with potential customers like an NVIDIA, like an MND, AMD, where they're saying for the next car of five years will supply this amount of memory chips. While they may not take the full advantage of the current high prices of memory, it does help to reduce a lot of the volatility because they're locking in, out of that, locking in, selling these memory chips throughout the next couple of years. Then I would say the final thing that we're seeing that's different from before, is that a lot of these companies are starting to work with their clients to produce custom memory chips. It's not going to be a pure commodity like RAM was in the past.