Kevin T. Carter, EMX ETFs

Description of it. But Burton was the co-founder and chief investment officer, and we hired some iShares people and built a direct indexing platform and sold that to Natixis in 2004. And about the same time that we sold the company, Burton had started to get interested in China and had started going to China and wrote a paper about China. And we had some investors that heard about this and called us and said, we want to add some China to our portfolio. And I said, okay, well, let me try to figure that out. And that was 20 years ago. And I'm still stuck in emerging markets in China trying to put together the best ways for people to invest long term. And as you may not know, but Burton and I partnered with Guggenheim initially on that effort and launched a

Number of China ETFs, including CQQQ, which was the first China tech ETF, which we launched in 2010. And then in 2014, I realized that all the things I saw in China with, billion new consumers getting their first computer, which was a smartphone, and then, doing everything online, I saw that that was going to spread to other countries. And Mercado Libre, which some of your listeners may know that the Amazon.com of Brazil was the company that I first noticed. And I said, well, this is a lot like the China internet companies. And that led to EMQQ, which is focused on the emerging markets internet sector and has been for the last 12 years. And now, of course, AI is the next tech wave. And it's also brought emerging markets really back into favor,

Because as it turns out, emerging markets are quite important in the AI world. And so we created a new brand EMX ETFs to build some AI related emerging market exposure vehicles. Yeah, and I want to talk about that. So you have a really great suite through EMQQ global, you guys, EMQQ, FMQQ, you and I talked extensively last time about INQQ. So just out of curiosity, like what was the, the reason or the thesis behind launching an entire new brand and not just put everything under the existing brand? Well, to be honest, it was it was kind of a coin toss. And it's not clear we won't merge the whole thing together. But the the EMX ETFs are actively managed. They're done in partnership with title financial group, which you certainly know,

Some listeners may know, but this is the white label platform that's the leading white label platform was founded by two of my, by good friends. When I launched EMQQ title wasn't in the business of white labeling. So, so there wasn't really a lot of choices 12 years ago. But so that's the reason for the new brand. And again, I think it's possible we'll, we'll roll them together, but we were eager to get moving and it, we set it up as a separate, entity within EMQQ global. Yeah, let's get right into Tigers, TGRZ, the EMX ETF, China AI, Tigers LLM ETF launched August 26th at a high level.

Can you talk about what this fund is and the problem that it's really trying to solve? Sure. Well, if you look at AI, the best way to, for people to get their arms around AI and what's happening is to, to use the five layer cake metaphor that Jensen Wong has popularized or the five layer AI stack. So if you look at what AI is, there's basically five layers. The bottom layer is power, energy, or more specifically electricity. The second layer is the chips layer. Now that that's a catch-all that includes semiconductors, memory, optics, server, all of the hardware, cables, racks, cooling, testing, all of the, associated parts of, of the AI machinery in the data center. And then of course the data center is sort of the third layer. So those, those first

Three layers are the infrastructure for AI. And if you're a hyperscaler, they're, they're an expense. this is one of the big, one of the, the several, um, questions that are, trillion dollar questions out there. One of them is how will the hyperscalers get a return on this investment? But if you're someone that is generating revenue from that infrastructure, you're, it's a business for you. You're the picks and shovels. And so right now, and, and, for several years that go for several years going forward, there's going to be a lot of picks and shovels needed. And, and this is where EM comes into play because as it turns out, well, the, the, the main AI race is between China and the United States. The people that are most important in terms of the

Hardware, the chips, um, are in Korea and Taiwan. And so, so those are the first three layers. Now on top of that, um, infrastructure, the layer four is the models, the large language models, um, which is what Jensen calls the magic layer. These are the real AI companies, if you will. We, we, of course, in the U S we have Anthropic and Claude, we have open AI and chat GPT, uh, Google and Gemini, and there's several others out there, but those are the big three, let's say, well, China is some of you, uh, some of your listeners know has their own models and they are different than our models. The main difference is that the U S models are all closed models.

The China models are all open weight models. And what that means is that, um, uh, users can download the model onto their own machines. They can tweak and customize the models, the weights, if you will. Um, they are just as good getting, nearly as good as our best models. And they're a fraction of the cost. In some cases they're, they're 90% less and they're getting used. People don't appreciate how widely used the China open models are now. And, uh, according to many estimates, about 80% of U S startups are using China models for the reasons I describe customizable on premises, own control, your own data, just as good as the expensive models and, uh, and very cost effective. So, um, and it's, it's not just startups. Airbnb is built on top

Of China models. Um, some people probably know about the cursor acquisition, which was a big acquisition that SpaceX made, uh, in the last couple of months that company's built on top of China models because of the reasons I mentioned before. So, so the tigers, it's meant to capture that layer. And the term tigers is used in China to describe, um, six specific models. So then in, in, in China, you have the, the tech, um, the historic tech leaders like Alibaba and Tencent and Baidu, and they, they have models and mainly Alibaba's QN model or QN model, um, which is the most downloaded model in the world in the last, uh, several months. Um, but then you have these six companies that are, that are described as the tigers. And those are, um, companies, all of which came from the same

Computer science lab. So Tsinghua University's computer science lab in Beijing is perhaps the best in the world. And all six of the tigers came from that lab. You have Z.ai, which is the largest holding of the tigers, which was the first large language model to ever go public anywhere in the world in Hong Kong earlier this year. You have Mini Max, which was the second one to go public also in Hong Kong earlier this year. Then you have Moonshot, which is the, the lab, the open model lab that created Kimi K3. Now again, for AI followers may remember in July, there was a lot of noise for a week, basically when Kimi K3 launched and it was nearly as good or essentially as good as our very

Best models. And again, at a fraction of the price, um, you also have a step fund. You have Bai Chuan AI, which is a healthcare focused, uh, model. Um, and then you have zero one dot AI, which was founded by Google's former China head, uh, Kai Fu Li. So those are the, the tigers, according to the official, there's no official list of these companies, but those six. Um, so we, the, our tigers, the TGRZ, we're not limited to just those six. We can own, um, Alibaba and others that have, large language models. And that important, that includes another important company that will be part of the tigers, uh, when it goes public and potentially via private investment beforehand, which the funds allowed to do. And we're

Exploring. Um, and that's, that's deep seek the model that really put China on the map. If you remember, it was only 20 months ago, January of 2025, when, um, there was a lot of things happening in the world back then. There was a, a re renewed interest in emerging markets and international investing by, investors and institutional investors. But then you had the drop of deep seek. And remember that the world, the investment world was already concerned about the capex that was going on and the amount of money that anthropic and open AI needed. And people were concerned about the payoff. And then this China model comes out of nowhere and they say ours is just as good. And it only cost us six and a half

Million dollars. And it had a significant impact. I believe NVIDIA was down 35% on this news. And it really threw into question whether this massive multi-trillion dollar spend was necessary. And so the reason moonshot or the reason deep seek isn't one of the tigers is it was actually started by a hedge fund. So the, the founder of deep seek was a math prodigy in China. And, uh, when he got to college, he was already working on quantitative investment, strategies using repurposed NVIDIA gaming units that he bought used. And then he started a hedge fund, which has done very well. And because again, it was a quantitative hedge fund, they had a lot of NVIDIA chips and he had a small team in the, on the side,

Develop this model. And now it's, it's become a major player and they just raised their first round of funding in the last two months, which was a $7 billion a round. And the founder himself put in 3 billion. So what you're going to see in the next six months. And in fact, it's timely because moonshot again, the model behind Kimmy, uh, they just filed to do an IPO in Hong Kong this morning or last night in Hong Kong, uh, maybe by the end of this year or early next year. Deep seek also is actively pursuing an IPO. So you're going to see these models, hit the, hit the tape, if you will. And then the other thing you're going to continue to see is this,

This competition between our leading closed models, Anthropic and open AI, which are both going to go public themselves. Right. And that's going to get a lot of coverage. Um, uh, but they're competing and they're competing in the United States. And you saw a couple months ago, there was, some talk from the incumbents about maybe these models, these open models are dangerous. Maybe we shouldn't allow these China open models. And the rest of the technology industry rebuked them. Satya Nadella quite, quite, actively said, look, that's not the way it works. These are great models. And, and that they were going to in fact, start employing deep seek themselves in their offerings. And that that's continued to, to, to develop. There was a letter penned by Jensen

Wong a few weeks ago during this, public discourse about this. It was, uh, Jensen Wong wrote it over a weekend. He got, I think 50 or 60 other tech CEOs to sign the letter. It was his first tweet apparently ever. And so this, this is going to continue this, this open model versus us closed model, um, race you this morning, you see the news about hugging face being acquired by Nvidia hugging face sort of the, the clearing cows, if you will, for these models. Um, and you've also seen reports that moonshot is in active discussions with all of our leading cloud providers for revenue sharing deals. So this story is really just getting started with the tigers.

Yeah, no, I'm we, you've covered a lot there. And one of the things, I don't know how much you are utilizing some of these models. I am an early adopter of all of them. I, I've basically, um, started just testing with a lot of these things. It's, it's actually interesting. Like the Kimmy K3 models, um, seem to want to, when you give them something to do, they seem to want to go and actually do it. It almost seems like our closed end models seem to want to keep your attention and like, they don't, yes, they perform, um, I would say better, but it takes more of your time sometimes. And that's just something I've noticed that some of these open models or some of these Chinese models, uh, seem to

Take less of your time and do more work rather than you having to like interact with this interface so much and so often. So, um, very interesting stuff. And I think it's odd that like, this is all just getting started, but let's just, let's get back to tigers. Cause I want to understand, um, some of the things that could possibly end up in this fun later on. So, right now is it's all models. Could you see as this, uh, or I should say all emerging market models. Could you see that investment thesis expanding as the story changes, or would that be a totally different ETF? No, we, we actually have, I, uh, we have seven different ETFs registered and effective. And, um,

Uh, the initial plan to be frank with you was to make, um, EMIQ, the broad emerging markets, AI, uh, portfile, sort of like an EMQQ for the, for AI. Right. And, um, uh, and that's still our, one of the things we plan on launching, but that's a full stack emerging markets. So, and, and again, it's, it, that, that's, that was the first, um, plan, but then our partners at title said, well, why don't we take that stack? Cause there's a lot of little pieces in that stack that are the models, for example, but there's also, um, uh, GPUs, the NVIDIA competitors that are coming out of China, which there's four, all two of which were founded by former NVIDIA

People. And one was founded by the former head of China's, uh, business for AMD. Um, there's the, the, um, optics and lithography, another layer where China sort of dominates. Um, so we have, we have full stack coming, but then we have pieces like the tiger. So there will be one solution that owns all the parts of AI and then several that are pieces of that layers of the stack, if you will. Just out of curiosity, I've had a couple of ETFs on the show that got, early exposure to SpaceX while it was still private. Is that something that's a possibility in China with some of these things like deep seek and moonshot that aren't yet live? Like, are you able to, through intermediaries get exposure there or do you just have to wait?

Well, per the prospectus, we are like a lot of other ETFs these days able to put up to 15% into privates and we are actively exploring that. Now we remember the tiger has just launched, a week ago. So, um, we'll have to grow the asset base a bit, but, but I think it is absolutely possible. And, and it turns out, having, um, been involved with China and emerging markets for 20 years. One of the best things I did 20 years ago when I got started was I, I hired about 10 Chinese MBA students here in the Bay area and had them work as analysts for me.

And I became very good friends with some of them. And some of them have now worked their way up, into senior roles in investment banking in China. And so, um, I, I know who to call, uh, once we're ready to, to put some money to work in that fashion. So we're absolutely looking at that right now. It's a little premature, but, but it has been part of our thinking from the beginning and why we wanted to get the, as narrow as the fund is now, we wanted to get it out there so we could try to get that, that pre IPO exposure to companies like moonshot and deep seek in particular. So as you said earlier, you're a, you're a buffeteer. You've always been kind of an

Active value investor first. How does that value lens actually show up in a fund built around such frontier technology like AI models? Well, I don't know that it does. I'm a businessman also. And, uh, I, to be very clear, um, the China, and this is probably the case with all large language models, including anthropic and open AI, but these, these are, these aren't making any money. These are not companies that are, uh, profitable. There's lots of questions of our question marks about if, and when they will be profitable, how profitable can they be? Um, there are a lot of people that I trust that have, um, and perhaps agree with as well, that there's going to be a lot of commoditization of the models and

That, um, it's going to be hard for them to have a moat and hold the pricing power. And, and so now it, I guess important in that calculation though, is it's the open AI, the closed models that I think are the ones trying to develop that moat that are the ones most at risk. The open models, I think are going to rely on, on other, um, revenue streams and ways to make money, um, with their lower cost model. So I think that this is not an Omaha investment, uh, today. It's certainly not the tiger's part of it. They could end up becoming, some of the most important companies in the world, but, and obviously right now, open AI and anthropic are two of the most important companies in the world,

But they're not profitable companies. And, and ultimately in Omaha investing is really simple. The reason companies have value is because they make profits for the owners of the company. And the only way to grow that value is to grow the profits. And these companies are growing their revenue. Hopefully they'll, they'll, they'll get into, cashflow and, and, earnings, uh, profitable mode. But right now I'd say this is a highly speculative, uh, part of the market. Now there are other parts of the stack, the AI stack where I think there is value. Um, and that's, um, right now I, people, one of the, one of the other funds that we've registered again, a very narrow fund is, is called the mag three, the, the emerging markets, AI mag three or

Emag and it, which hasn't launched, but in, cause I, what happened with, with me, I obviously have been focused on tech and China and emerging markets for 20 years. It was very natural for me to, to, be early and thinking about AI and how it was shaping things. And the thing that I saw that every investor saw, and I don't think they quite appreciate how big it is, but emerging markets in the last 20 months since January of last year, they've outperformed the S&P by a decent amount, but there's basically three companies that are making all the money and providing all the return for investors. And those three, the mag three TSMC, Taiwan semiconductor, which is perhaps the most

Important company in the world. It may not be the highest market cap, but they make all of the chips for basically everybody, including Nvidia and Apple, et cetera. So TSMC. And then you have the memory leaders, Samsung and Hynex. And those three companies combined, um, are now 33% of the index. it's the most concentrated broad market index maybe ever. There's single country, uh, ETFs that are highly concentrated because they don't have that many companies, but, but this is, this is very concentrated. It's a significantly more concentrated than our market. Um, and, uh, those three companies have also provided 80% of the return of emerging markets. So emerging markets have done great, but it's been almost exclusively those three companies that have driven it. And it's wreaked havoc in the emerging markets, both positive

Havoc and negative havoc. So consider, consider this since last January, again, 20 months ago, the MSCI emerging market index, which is what iShares emerging market ETFs track is up. I'm going to round these numbers, but call it 65%. The Vanguard emerging market ETF, which is the largest, it tracks the FTSE index. It's only up 30% in that same period. Why? Because the FTSE index doesn't own Korea. And so people have known about that and it never seemed like it was that big of a deal. Korea wasn't that, large of a market relative to the other emerging markets, but, but that one difference has led to a massive, uh, difference between the Vanguard emerging markets ETF and the other MSCI related, uh, funds. So

Now the other thing about the MAG three though, you, you, we have all these concerns about, um, the CapEx, right? How are we going to make any money? The hyperscalers are now cashflow negative. They're tapping the debt markets that this is, this is that one of these trillion dollar questions I'm talking about, right? Well, on the flip side of that, the MAG three are making so much money. It's hard for one to get your arms around. So between 2026 this year and next year, 2027, those three companies are expected to make 965 billion, 965 billion of profits. How much is that in the entire history of Berkshire Hathaway.

They have not made that much money in Amazon's 30 plus years of existence. These companies are going to make three times as much as Amazon's made ever. Right? So now the valuations go getting back to Omaha. If you look at those companies today, in particular, um, Hynex and Samsung, their PE ratios. What, what do you think the PE ratio is for Hynex? Let me ask. I don't know. I'm going to guess. I'm going to say, uh, 58, three and a half. Oh, that's compelling.

Collectively the, the MAG three have a PE ratio under five. And I, and, and the reason there's multiple reasons. that there's, um, this fear that these are cyclical businesses, right? And they have been historically, right? Micron in the U S has been a, boom and bust kind of operation. And the question though, is, is this a cyclical thing or a secular thing? And I think you'd have to say it's a secular thing. Now there are, there's definitely competition coming, right? You, you see, Amazon selling chips, Google's, a lot of people are making, getting in the chip business. So there is that element. And then there's a question about whether there will be new forms of memory that are better suited for AI. But as of right now, I,

I think people would be a little surprised. I just started asking people what they thought the PE was last week. And first reaction I got was 90 is what they thought the PE was, but it's three and a half. So Kevin though, with those. So I guess the question is just staying on these, these three, is that where the true bottleneck lies in this whole process? Because memory shortage and like, in your opinion, is that what could slow this whole horse down? Um, well, it, it, it is already slowing things down. it's, it, it, it's not slowing down the hyperscalers, but it's, it's wreaking havoc amongst smaller players. there are a lot of companies that are small companies, not public companies necessarily that are basically going out

Of business because that input cost of memory, in addition to being up several fold, they just can't get it. It's not available. there's a shortage. So that is a bottleneck. I think the, the biggest bottleneck that we may see is the power at the electricity. And we are moving ahead at a breakneck speed, but the power supply is not. And the, all of the, turbines and generators are sold out for years. They're resorting to recycling old retired airplanes and getting the engines off and turning those into power. Meanwhile, China has three times the power we have, and they have incredible velocity. Last year, China added more power than any country in the world has other than the U S and China. And in the last four years, they added more power than we

Have. So they've got three times the power. It's a lot cheaper. And I suspect that we will have a bottleneck in the U S there's obviously efforts to increase efficiency throughout the stack, but it just seems to me that that's a problem that's coming fast here. But it's, if you compare and contrast to China versus U S China clearly has the lead at the power layer, we talked about the shortages at the chips layer, but there's also, I think going to be a shortage and a bottleneck at the optics layer because it optical, basically what's happening is copper wires are getting replaced by optical gun accents. And there's a lot of reasons for that mainly that's better. And, and that doesn't leak as much as the copper. Um, so that might be another bottleneck, but I think power and continued

Shortages in the various parts of the chips layer optics is one that seems to me, uh, might be a problem. Um, and I should also mention there was a, there was talk in Washington in the last month or two about possibly banning the, the Chinese optical suppliers, which I was pleased. I saw the CEO of the largest U S optical manufacturer, but he was asked about this and he said, don't do it. Don't, if you want to see shortages, take the Chinese cause they're the dominant supplier right now. And so if you take that out, then even though this gentleman is running the company that would benefit the most, he said, no, don't, don't do that. Otherwise we'll have a real bottleneck in that part of the stack. So.

So Kevin, just as we close here, just one quick question, like how should an advisor view this tiger's ETF to me, it seems a little bit more like thematic exposure rather than pure emerging markets. Like you have that exposure with your EMQQ suite. This seems like pretty thematic. Is that the right way to think about it? Oh, it's absolutely thematic. again, it's, I think that it's sensible for advisors to be wanting to figure out how AI is in their portfolios. And again, the emerging markets, AI, strategy that we will launch soon enough that will capture that. That's really a product made for advisors to just capture the AI theme within emerging markets. I think that these single layers, whether it's the mag three or the tigers or the

Optics, these are more for active investors. And, and I'm sure there are advisors out there that are a bit more tactical, but for strategic allocations, I don't think this, that the tigers really has a place. I think those companies have a place, but within a broader EMIQ, broad emerging market AI strategy. Well, Kevin, I always appreciate you coming on. I learned something every time, but before I can let you go, where can people learn more about the fund and EMX ETF? EMX ETF.com is the easiest way to find out. Well, again, thanks for being here with me.

Okay, Brad. Thanks again. Great to see you. Bye.