Under the surface, key parts of the A.I. trade include memory as well as phonics and optical networking. Memory is a specialized semiconductor components that feeds GPUs with Sandisk being a prime example of a memory stock. Now, meanwhile, phonics and optical networking bypass electricity to connect massive clusters of A.I. processors. Now, joining me this morning is Yuri Kajimari, CIO of Tema ETFs. Yuri, thanks so much for joining us today.
Thanks for having me.
So let's go ahead and start with a breakdown of the AI trade. Memory has a huge winner in the first half. Take us through the memory part of that specific trade.
Yeah, absolutely. I think the way to think about memory is, what are some of the key bottlenecks as we progress through this AI revolution? Obviously, the key component were these NVIDIA GPUs. But the reality is that for AI to deliver on its promises, particularly agentic AI, we need a lot of memory. And we need memory across all of the different types of memory that are present in the data center. So here I'm talking about high bandwidth memory that sits right there with the GPU to perform the key tasks at the highest speed, but also NAND, which is a colder, slower form of memory, but it's used for storage. Essentially, what we're seeing is that computing power has exploded, but memory has failed to keep up because of capacity constraints and is becoming an important key bottleneck within the AI ecosystem and build out.
And let's go ahead and talk about the investor demand specifically for this area when we talk about the optical aspect of it.
Yeah, absolutely. Look, I think investors are understanding that for us to deliver on the promise of the AI revolution, and by the way, it's absolutely happening out there. Token usage is exploding. Companies are adopting AI. Businesses are doing that. If you listen to Amazon, Microsoft, any of these companies, they're talking about essentially insatiable demand for some of these AI products. And for that to be delivered within these data centers, as I mentioned, the computing power of these data centers keeps growing. but it's starting to hit bottlenecks. And some of the bottlenecks are related to memory, but also to the ability of moving this data. So you need to move vast amounts of data over distances within data centers, connecting chips, different server racks, different data centers together. And simply put, copper, which has been really the piece of technology that we've used to connect and transfer data, is hitting its physical limits. And so we're going to enter a phase, and we're already in it, where optical and photonics are going to replace copper as a key means of transferring data. Simply put, this allows you to use the full spectrum of light to transmit data. You can transmit data much faster. And this is going to be a key part for the rest of the infrastructure to keep up with how fast GPUs are expanding.
So you guys recently launched two ETFs to track the AI trade. So let's start off with the memory fund. Tell us about this fund and its exposure.
Yeah, absolutely. So we're really happy to launch a couple of weeks ago our ETF called DISK, D-I-S-K. It focuses squarely on the memory opportunity. As I said, it's one of the most exciting parts, partly because there's this clear, insatiable demand from the AI build-out, but also there is a very, very, very acute shortage of supply. If you listen to SK Hynix, Samsung, Micron, these companies are saying that memory is sold out certainly for next year and now we're looking into 2028, 2029, potentially for these shortages to continue. This is driving prices up and creating a big profit cycle for these companies. You can't deliver agentic AI without memory. It's a key component of it. Now, DISC is different to some of our competitors, because we focus particularly on things like NAND, where we see a lot of opportunity. Price increases of NAND are our pacing, actually, DRAM. And we think there's going to be a push, because DRAM is so expensive, to offload some of this data onto NAND, which is cheaper to deliver per gigabyte. Another part of this is focusing on some of the supply chain. So interesting companies that you may not have heard of out in Asia that are going to be a part of this ETF. So we think this is a compelling investment, even for those investors who are already invested in memory, as a complement focusing on NAND and some of the supply chain out in Asia.
So on the flip side, you guys also launch LAZR. So expand on this and its pre-IPO position in Anthropic.
Absolutely. So we're, again, also very excited to launch LASER, L-A-S-E-R, which is the Tema Photonics and Optical ETF. The focus here is on that opportunity which I described. Copper is hitting its physical limits, and we're going to need to progressively replace all of the components within data centers to deliver data through optical means. Using the full spectrum of light increases bandwidth, increases speed, and allows data to move more freely in the data center unlocking more opportunities with AI. The fund itself is a pure play exposure to this. So you've got companies like Lumentum, which people have heard of, but also businesses out in Asia that are delivering transceivers, optical components, really part of this whole AI infrastructure revolution. We're also really pleased to have as part of this and as part of the AI infrastructure build-out a pre-IPO stake in Anthropic, which is a key participant within the AI infrastructure system and a key buyer and demand source for a lot of these optical components. As many of you know, Anthropic is pretty much the leading LLM out there. It's clear that the company's revenues are scaling very fast, and they're at the forefront of AI demand as an AI revolution.
So where do you think that the AI trade specifically goes from here? And is there room to continue running?
Look, I think for investors, especially in the last couple of weeks, there's been this shakeout in the AI trade and there's been a concern about the returns of AI from all of this capex. I think what I would point investors to is a couple of things. The price of intelligence was always going down. And as the price goes down, usage continues to go up. And so we watch very carefully usage and the revenues from the large labs. And these continue to scale really pretty much in an unprecedented pace. So we think the ROI is there, and that's what the big hyperscalers are telling us. So that means there's going to be more and more investment in the AI infrastructure for the years to come. And the key thing for investors is to focus on where are the bottlenecks? What are the key hurdles where we can improve the capacity of these data centers? For us, that's memory, which is a key important part of this. And it's also photonics and optical, which is a key part of carrying that data all the time.
Awesome. Well, Yuri, thank you so much for joining us again. Yuri from Temai ETFs. Thank you.
Thank you.
