AI Data Center Spending Fuels Massive Demand for High-Bandwidth Memory Chips

Joining us to explain this wild ride is Dave Mazza, CEO of Roundhill Investments.

Dave, good morning.

Thank you so much for joining us.

So let's start off by looking at this is the Round Hill memory ETF and that did break records and hit over $10 billion in assets.

6 weeks, but we have to keep in mind that memory is also a notoriously boom and bust industry and is also cyclical.

So do you think investors are so blinded by AI hype right now that they're forgetting how brutal the downside of the cycle can also be.

Well, I think it's interesting.

Investors are really recognizing that there's a few bottlenecks of the AI with memory chips actually being one of the largest bottlenecks in the market today.

And the reason for that is as the data center build out spending continues to grow from $500 billion 600 dollars, $700.01 trillion dollars of spending, where does that money go?

Well, it goes to chips.

And in order for the most effective, powerful AI GPUs to actually run effectively, you need memory chips, but there's actually a massive supply constraint in the market today because only a handful of companies really make high bandwidth memory and digital random access memory chips.

And so when we introduced the DRAM ETF.

Recently, investors have really gravitated toward that because they're recognizing that this theme has the potential for there to be a long term rerating of these companies because as you stated historically they were viewed by the market, subject to boom and bust cycles, but that's began to change with this AI.

And Dave, the new fund is pulling in billions quite quickly.

One of the fastest growing ETFs in history, but nearly half of the portfolio is tied to South Korean equities.

South Korean names like Samsung and SK Heinek and you are joining me on this Wednesday, which is Nvidia earnings day, but Samsung and its largest labor. are locked in negotiations to avert a supply disrupting strike fueled by workers demanding a larger share of the company's AI-driven profits.

So with so much exposure to overseas governance as well as potentially affects rates, is it just safer for regular investors to buy an American company like Micron directly.

What are your thoughts here?

Yeah, I think investors certainly have gravitated toward Micron and some of the storage names like Sandisk, Western Digital, and Seagate, which are all in the Round Hill Memory ETF.

But what's interesting is if we're just focused on Micron, you're missing out on two of the other largest memory companies, which is Samsung and SK Heinx, which historically were very difficult for US-based investors to access due to the fact that these companies do not have US ADR.

And so to speak on the, excuse me, the Samsung point, yes, I think we've seen labor negotiations break down.

That's not actually going to help the supply side, particularly if workers go on strike and not have the ability to work in their factories.

So it's a really interesting setup because the underlying market for memory chips, so the price of them, has skyrocketed, and there's actually an expectation that that will.

Continue to be the case because especially if Samsung goes offline here, the ability for these companies to create new factories takes 3 to 5 years to create a new fab facility.

In fact, the equipment that even if they open the factory today, the equipment that they would need is sold out to actually make those chips through 2026.

So this is one of the reasons why there's kind of a long term tailwind for memory, even if it's, it's going to be subject to some volatility in the near term.

Yes, and Dave, I know you're entrenched in the space.

So can you tell us why the hard drive, as well as the storage side of the business is acting so differently compared to the pure memory chip side right now?

Yeah, I think first and foremost, you know, these are, these are smaller companies.

I think Micron, SK, Hein, Samsung, these are multi $100 billion companies.

In fact, they're expected to produce some of the largest profits around the globe in case, you know, investors just aren't as familiar with them, it's because they're so focused on making these high bandwidth memory chips.

The storage names are a little bit different, um, and so the way I think about it, it's almost like a kitchen where the GPU, the Nvidia chip, acts as the head chef.

The memory is the miseen plot where all the information, all the ingredients can be taken very, very quickly to make a dish.

And then the storage side, um, which is your, your, uh, sand disk, Western Digital, that really is almost your, uh, your refrigerator or freezer.

And so as these AI models move just from training to inference and then agentic AI, all of those components are needed just like all of those components are needed in a kitchen in order to to run effectively, um, to serve, to serve meals to diners.

Um, so that's why one of the reasons why we combine the memory names and, and some of the the more specific storage names in the DRAM ETF.

And finally, Dave, before I let you go, based on some of the data as well as pricing that you're seeing, should everyday investors out there keep chasing this momentum, or do you think it's time to literally take some chips off the table?

We, we, we've certainly seen a historic run, coming out of the market bottom.

Um, so these stocks actually have been sort of the top performers, um, as, as we've seen the market rebound from, from their late March lows.

Um, so I think the moves again have been pretty historic in many ways, even if you look at a year to date, one year numbers quietly under the radar, um, certain investors have gravitated towards the bottleneck trade.

But what's interesting is that these companies still trade at significant low price to earnings multiples compared to other semiconductors compared to the broader technology sector.

So as these companies quarter after quarter continue to produce the revenue numbers that we expect them to do, we think there's a potential for a long, long term rerating of these names.

So while the gains may not be as easy as maybe we've seen over the last 6 weeks, we do think there's a significant tailwind for there to have, you know, robust, strong performance on a go-forward basis.

Well, Dave, we will have to leave it there for today, but always great talking to you.

Thank you so much for joining us this morning and thank you so much for sharing all of your insights.

Thank you.