#1035 - What's the Best Way to Play AI? ft. Joe Zhao

0:00Hey, visionaries. Today's episode is brought to you by Polkadot, a leading layer zero blockchain with over 2 ,000 developers. It's a network protocol that allows arbitrary data, not just tokens, to be transferred across blockchains. Listen to what Polkadot creator Gavin Wood tells Raoul about Polkadot's coming jam chain, short for Join Accumulate Machine. So what we're doing is we're turning what used to be the Polkadot relay chain built for a very specific purpose, right, to secure and relay messages between separate blockchain ecosystems. And we're turning that into something much more akin to this like world computer, this like kind of ubiquitous multi-core single-turn virtual machine.

0:45Learn more and join the community now by going to realvision.com slash polkadot.

0:59What's the best way to play AI? Hi, everyone. Welcome to the Real Vision Daily Briefing. I'm Maggie Lake. With me today is Joe Zhao, partner with Millennia Capital. Hi, Joe. It's great to have you back on. Hi, Maggie. It's good to see you again. So I'm really looking forward to this. And I had to really resist to start asking you questions before we came on air because we were just off camera for a second because you're right in the thick of the technology space. And in particular, I think you're focusing on AI. So it's a really interesting time to catch up with you. So I just want to get a sense of what you've been up to first.

1:33Then we'll talk a little bit about big picture and macro and how some of the AI investing trends are colliding with how it's working with the global macro backdrop we're in. But what's happening in the world of tech investing? What's the mood? What have you been up to? Fill us in. Well, thanks, Maggie. And it's great to be here and saying hi to everybody. Last time I was on was last year. And I've been nonstop just working, traveling, working, investing, raising, managing and different contexts. Millennium is an investment firm in New York City. We also have some partners in operations in California and Miami but mostly based out of New York City.

2:13And right now we invest mostly only in private markets. Eventually we may do a public market fund but right now only private markets. And especially within that context we invest in growth and later stage of venture backed startups in technology and we were doing a bit of AI even before AI got really, really high the last couple of years. We were investing in technology software, cloud fintech, AI in 2021 and prior. But I think after the GPT moment at the end of 2020, early 2023, that was sort of the validation. And since then, we've also added a number of other investments in AI to our portfolio.

2:57amongst them are Cohere is one of them. Lambda Labs is another one that we invested in. Stability. And there's a couple of other ones that we've invested in that we haven't announced much yet. But like we are, you know, I've been, I'm in California this week and we've been seeing a lot of AI founders, entrepreneurs, investors. And so, you know, to your point, I'm really, really fortunate to be at this point where we're sort of out of, I've been sitting on the front row in seeing the AI deal flow and I've been able to invest and in that context, I've just learned a lot about, you know, the dynamics of about the macro trends.

3:39I'd love to share something that with the audience here. Yeah, absolutely. And we so appreciate that because it's a peak at a world that, you know, most of us don't have access to. And so just to be, I just want to underscore a couple of things you said. So, you know, Millennia is for institutional investors. You don't have a public fund right now and you are some of the trends and things we're talking about are happening in the private market. So you're seeing it early. So this is really you being generous and sharing what you're seeing on the trends for us as opposed to naming names that the rest of us have an opportunity to invest in.

4:11We're not there yet, but understanding what's happening, especially in the backdrop where a lot of people are saying, oh, it's hype or we've reached peak hype. Seeing where the smart money's going is really important to the rest of us. So that's super important. And I'll also say, though, the fact that you are investing in private markets, if there's something that comes up that you're not able to be specific about, let us know because there's a lot of compliance and regulatory issues with this, folks. And so we totally understand that and appreciate that if that comes up. But talk to us a little bit about that issue of hype.

4:46You know, what are you seeing? Because it felt like it was NVIDIA to the moon. And anybody that mentioned AI in a conference call, there was just sort of blindly money flowing into it, or at least that's the narrative. And now everybody's kind of like, well, what does this really mean? What are the realistic applications for this? How do we separate between the hype and the real opportunity? What are you seeing in terms of that? Yeah, so I'll start by saying that, like, I 100 % believe that AI is real and that it's going to really change business, society, consumer lives as we get over the coming years, 5, 10, 15, 20 years.

5:28I think we tend to underestimate the impact these technological shifts can have on society. In 1994, we thought internet was going to be around, but now it really changed our lives. Now all of our banks and money are just on a phone app, right? but I think in the next 10-20 years, a lot of that will change again. And so, and obviously in that context, when something is so real and transformative in the market, there's always going to be noise and substance. So I think a part of the job as an investor, a good investor, whether you're retail institutional, is to try to think through what is noise and what is green.

6:11And if you're able to do that, And that's part of my job is to detect the real AI companies from the not so real companies. And there's a few things I can, there's a few sort of things I look for I can share. But I think in that, if you can bet on the real AI trends, the real companies, and if you can do really well as an investor. And so in the private markets, like I, you know, we've been, you know, our clients are mostly family offices, institutional, some high net worth. And we've been able to invest in some of those private companies on behalf of them. But with that said, like one can still play this AI trend by investing in public markets.

6:56And so, but obviously a lot of the innovations happen in private markets before they get bought out or they go public. and so happy to kind of share some of the things we're seeing in private markets. Yeah, I'm interested if you've been able to determine where you see... So what are the parameters of what makes it seem real, right? It's not even what they're doing. It's like, what are the sort of hallmarks of something that you feel looks like a real opportunity as opposed to noise? I'll start with a couple of things that really excite me and I can kind of go down to the infrastructure that support that possibility.

7:36And those are things we're investing into. So I'll give you an example. Like today, I think we just saw on the news that like OpenAI launched something where it's like there's AI being trained on words, AI being trained on images, and there's AI being trained on sound, right? And so there's a few things that really suck. So for example, like we now have the technology to support a device that can listen to your body for sounds and tell you how your body's doing. There's no way my human ears could possibly do that. A doctor has to do that with the devices. I don't want to go to a doctor's office every single day, but in theory, we have technologies available to consumers in the coming months and years where the device with the AI can listen to our body's sounds.

8:22And what gets me excited is I think in the future, like we can have a device with the AI that will smell our body and be able to make a judgment about the state of my health. And so that's two applications. And I assume everybody's played with GPT. It edits my emails, reviews my legal docs. That's my financial models for me. Helps people code. I've done that. I've tried that before. And now with Sora, you can basically make a movie out of just the prompt. Right. And so we're in the first to the second endings of this AI revolution super cycle. And imagine if the AI can do these things in like, you know, in the first in the first endings of its development.

9:06Imagine what it could do once we go to like half time and we go towards the end of the game. Hey, everyone, we're going to take a quick break right now to hear a word from our partners. We'll be right back with more of the day's top analysis on the Real Vision Daily Briefing. Have you ever wanted to trade Bitcoin but haven't dared try? With Plus 500 Futures, you can trade crypto without the hassle of opening a wallet. With just a few clicks, you can register and start practicing with their free and unlimited demo. See a trading opportunity? You'll be able to trade it in just two clicks. Feel ready?

9:38You can move to real money with as little as$100 once your account is approved. And the great thing is that in addition to crypto, Plus 500 gives you access to a wide range of instruments. S &P 500, NASDAQ, gas, and much more. explore equity indices energy metals forex and beyond with a simple and intuitive platform you could trade anytime anywhere experience the fast accessible futures trading you've been waiting for with plus 500 with over 20 years of experience plus 500 is your gateway to the markets visit us.plus500.com to learn more trading and futures involves the risk of loss and is not suitable for not all applicants will qualify plus 500 it's trading with a plus

10:23Yeah, I think it's actually hard for most of us to imagine it, and that's part of the challenge of this. So are you looking at the use case that you like, which, of course, if we think healthcare, aging population makes sense, or is it more, when you say the application, is it more of a platform that can deliver it to the consumer? Is it going to be a med device, just a med device superpowered on AI? How are you actually walking through when you're trying to vet people who are looking for you to invest? So AI is a platform in my view. On the platform will be many, many new companies being born with many, many use cases to be discovered.

11:09I think I read a study that showed that we as an industry have already identified hundreds of use cases using AI. And the ones I've listed are just two things that I find personally fascinating. But you're going to see AI in everything from business to military, to government, to tax collections, to audit, to driving. So I think it's fair in my book to say that the possibilities for AI is limitless. and why do I say that? Think about it, right? A lot of the things we have in our society that are of value today, I'll give you a very simple example. Coca-Cola is a formula that makes that drink that makes it so delicious, but that formula came from someone designing it from their head.

11:58But if you can replicate that brain, you can design any formulas in the future that will be designing better, more tasty formulas than Coca-Cola, for example. So AI is a platform that will give birth to numerous possibilities that we can't even imagine. So how do we know what... So that's the societal impact. From an investing point of view, how do you pick the winner? Is there a moat around that? And if everything's going to be transformed, if everything's AI, then what's the difference? So that's the great question. And so now the analogy I'll give is we're at a point where it's too early to say which applications will be the most transformative in the next five to 10 years.

12:47I don't know. I don't think a lot of people will know. We have a thesis, but one thing that we know is if that's where the industry is going, then what are the short bets that we can make that will benefit from that sort of growth in the industry? And so I think that's why investors, smart money in the private markets have been investing in the infrastructure layer. That supports that possibility. So now what does that mean? I'll give you, let's just, let's say, let's imagine there's a stage, a platform. So on it will be many applications being born. Below that platform, what are the pieces that make up that platform?

13:27So one layer, there's really three to four things that make up that ecosystem of infrastructure. The first one is the large language models. So in that industry includes companies like OpenAI, Anthraffic, Cohere, and there's a few others. There's another piece of that, which is the data, because you have to have the data to train the AI. And so the data owners are the big tech companies like Amazon, Microsoft, Apple, Facebook, Instagram, et cetera, and what have you. Piece number three is sort of the chips and the chips, the chip sector. So that's like NVIDIA and the Intel. And then the fourth bucket is sort of the glue that holds all these things together.

14:17And so I'll give you one example. There's a company called Scale AI. It's very popular. It's a private startup. And what they do is they clean the data. So now how does that work? An AI is basically a reflection of a data set on which it's trained. And the data set is... So to train an AI, you have to have a few pieces. You have to have the data. You have to have the algo that trains the data to train the brain. You have to have the chips on which electricity trains the data to be intelligent so it goes into the brain. And that's the AI infrastructure. If you look even one layer below that AI infrastructure, how are these things made possible?

15:05You have to have electricity, right? Because AI is very energy consuming. and so you have so power utilities are a couple of really good investable teams behind that you also have to have like real estate for data centers so now to summarize it you have to have good data centers that are placed strategically to allow for two things one is it's stored in a good climate where it's got access to cold air so it cools down the heat of the chips and And they have to be placed strategically where there's low latency between the transmission of data between the population centers and the data centers.

15:47You have to pick locations for data centers where there's not a lot of natural disasters. So that's one thing. On top of the data centers, you would have a lot of chips, NVIDIA chips mostly. And you have to have a system whereby you cluster the chips together so the chips are super powerful. and you have to have a cooling system to cool the chips so they don't become too hot. On top of that, you have to have the data that go on top of the chips, electricity, and then the large language model, the algos, to train the AI. Only then do you have a working model that will allow developers to train the AI.

16:26And then so now, no matter which way the AI goes, whether we're going to be building AIs for healthcare, for defense, for tax collection, audit, and self-driving will all be running on this infrastructure. So what the smart money, I guess what the money is doing in private capital is we're all investing in the infrastructure, thinking that no matter which way the AI goes, this is the casino that's going to win the game. So we want to be investing that. And then we'll let the market play out. Yeah, that makes a lot of sense because that's the foundational layer. And as you say, it's too early to try to figure out what's going to be sort of the killer app, so to speak.

17:08And we sort of saw that with the internet, right? I mean, once the sort of foundational layer of the internet was there, then apps came and went, dot-coms came and went, but some of them blew up. So let's just interject a little bit about the macro before we get a little further in the weeds here, the macro backdrop. So we've had a fair amount of volatility. We saw an economy that was running hot and everyone backed off all these Fed rate cuts. Now it looks like it's kind of rolling over again on employment. What do you see happening on this sort of economic backdrop? And does it matter to the investments you're making?

17:46Sounds like maybe not. Yes and no. What I would argue is if the macro backdrop isn't this tight, I would imagine the AI cycle will be even bigger. Because a lot of the serious money, the pension, the sovereign wealth, the foundations, the LPs, the allocator, the family offices, they're parking their money in money market funds, getting 5 % risk-free. And they're making 8 % to 10 % in private public credit, 12 % to 15 % a year in real estate, private equity, and private credit. So they're parking their money in that part of the market. But as soon as the Fed eases financial conditions, the money will be moving back from the fixed income to the equity markets.

18:31And that will probably fuel even a larger rally in equity markets across NASDAQ, S &P, and probably private markets. So I would argue that if the macro backdrop were easier, the innovation cycle, the investing cycle will be even bigger. But to your point, there's a lot of activities in the AI landscape right now from both private capital and from public capital. So, for example, the federal government has been investing in the CHIPS Act to improve semiconductor resiliency. So they're building what chips manufacturers in Arizona and other states. So that's public spending. Then on the private side, I think I just read an article that said that last year in private markets, I think for$50 billion were invested in the startup space.

19:25and we're not even talking about the amount of money that went into NVIDIA and Microsoft. And so what I would say is there's a lot of investments happening, public and private, in AI, and in a very difficult macro environment. I would argue that when the macro gets easier for equity investors, there'll be even more investments going to AI. Yeah, that's a great point. It'll push people a little further out the risk profile, especially, as you say, some of that more conservative capital will be more willing to go into that area, which is super interesting. So we have a couple of great comments. So when you see something, I know it's difficult to comment on individual names, but when people are worried, can we use the same valuation metrics when we're faced with an opportunity that's hard to quantify?

20:19So we talked about infrastructure. We know NVIDIA had huge gains. A lot of people feel like that's overvalued. I don't know if you have an opinion specifically about that or whether you can speak to it or not. If you can't, that's fine. But do we need to rethink it or do the old rules still apply and we're going to get this big move and then settle down as competitors come in? Should we think about it the way we have traditionally? Yeah, just analyzing Nvidia stock as a bystander, trading at what, 20, sorry, about 30 times forward earnings is not that expensive relative to the total adjustment market.

20:54I think adding the fact that NVIDIA basically is a monopoly and privately when you talk to developers and the feedback you get is like, NVIDIA's chips are just way better than other vendors' chips. It's kind of like, the analogy is like, hey, if you can drive up from New York to Boston, right? And you can either take a plane or you can take the bus and you're maximizing for timing. What would you do? We're going to take another quick break to hear a word from our partners. We'll be right back with more of the day's top analysis on the Real Vision Daily Briefing.

21:34Yeah, exactly. When your investors are breathing down and you want results immediately. Or it's like, hey, you know, we're driving from New York to California, which I've driven a couple of times. And one is you drive four to five days without sleeping. One is you fly for five hours. Yeah. Yes, one is a bit more expensive, but it saves you a lot of time. Yeah. So I think that's... I think just as a buy-seender, NVIDIA seems to have the best product and has a huge market share. And I don't think that about$30 for earnings is that expensive relative to the total juggle market here. Yeah, that's why I asked.

22:10We have to think about it in different terms. so the the energy component is interesting right so we have a question um from ralph it's a two-parter but there's an idea that bitcoin mining facilities will flip to being ai data centers that's already happening yeah yeah there's a private startup called core weave and by this disclosure we're not we're not in the company um um which is why you know i could probably talk about it but there's a a startup called core weave and they just raised a massive round back by like numerous leading players including nvidia and um the genesis of that company was it was you know until 2021 2021 it was a big um uh gpu uh service company like chips service company cloud company where um they they they where they worked with many many crypto miners.

23:06And then after crypto cooled down in 2021, in the last couple of years, CoreWeb has been repurposing its services to serve the AI developers. And it's growing massively from what I'm hearing. I think last week they announced around the week before. And so you're definitely seeing some of that crypto infrastructure being repurposed to support AI development. Which is interesting if you're in the crypto space. It's like a little bit of competition for resources when it comes to something like energy and some of the infrastructure, which is a super interesting point. Is there any other kind of relationship between crypto and AI?

23:47Because the second part of Ralph's is that the co-founder of Palantir said for AI agents to coordinate with incentive systems, they're probably going to use crypto. There's some truth to that. I think like even when I was on other shows with them, at Real Vision I was asking. Even a couple years ago, I was making the point that blockchain is a unique technology where it's really, really special. But the question is, how do you use blockchain in a way that can immediately impact their lives? And I think one of the use cases is you could potentially use blockchain to regulate AI. And I think, on the other side of AI is all the concern about, hey, what if this AI gets out of control?

24:30And so instead of having humans to like regular AI, why don't we just draft our rules on a blockchain and have that regular AI? That way, the entire planet can be asleep and there's something working in the background of the regular AI. So that's theoretically where I think we're going and what people are thinking about. But I think we still need good entrepreneurs to figure out how to build a business around that. Yeah, exactly. And a lot of people paying attention to that because it's kind of the ethical piece while everyone's chasing the next unicorn that you need sort of good minds on. So I think this is a great question, Mark.

25:11This was my next question. You read my mind. Joe, do you think AI will be a fertile investment landscape long term or will the mega caps swallow it up and monopolize it all? So this is where the example of the Internet is really important, right? Because, yes, it was this transformation, but it's kind of owned by a couple of big companies that are, at the very least, the gatekeepers to most of it. So do we see that repeat, especially given some of the need for data and the fact that those companies, because of the last revolution, sit on a lot of that data? What's your thought about that? Yeah.

25:50So, you know, if you go back to my earlier point about what is the picks and shovels of AI, the infrastructure, right? There's like really four pieces. It might be one is the algo that trains the brain. One is the data. One is the chips. And the last is like the ancillary service, like clean the data set. Well, guess what? Public companies already are big players in that infrastructure. What does that mean? NVIDIA is like a monopoly in that chips market. And the big things, the big tech, mega tech stocks, the companies are huge owners of data, right? So for example, like Apple has all of our data.

26:35Microsoft has all of our enterprise data. Amazon has a lot of data. So has Google. And so in fact, I mean, that's why like GPT has been so successful was because GPT, OpenAI has had access to like Microsoft. resources. And I've had access to Google's resources, Amazon's resources. And Facebook, sorry, Meta's llama has had access to Facebook's data set, sorry, Meta's data set, which includes Facebook, Instagram, WhatsApp. And so there are ways to play the AI market from a public market standpoint. And just, I think, you know, and there's obviously, there's a part of that foundational layer, which are only available in the private startup space.

27:16But eventually, like, for example, Microsoft was effectively acquired. Sorry, let me rephrase. Microsoft, basically, Inflection's team has joined Microsoft. Inflection was one of the leading large language model startups and a bulk of their team has joined Microsoft. So that's an example of already a public player already absorbing some of the private talent. And I think that probably won't be the last. And so I think what I'm saying is to take a step back. Yes, we, the VCs, are playing. We're investing in the AI industry via the startups, but there are ways for public market investors to invest in AI through the public market companies.