Why Nvidia, Microsoft and Google still dominate the AI trade

Foreign.

Welcome to Trader Talk, where we dish out the latest Wall street buzz to keep your portfolio sizzling. I'm Kenny Polkari coming to you live from the Yahoo Finance headquarters in the heart of New York City, a global hub where deals are made, fortunes are built, and the next market move is always just around the corner. Coming up, I'll share my big take on AI. And I'm gonna sit down with Tsuro Capital Research analyst co William Lee. And I'm going to share my pasta alla norma recipe with you as well. Now let's jump into my big take. There's a lot of noise right now around artificial intelligence. The headlines hype it up. The new gadgets, chatbots, image generators, and all the buzzy AI apps landing in your app store. It's easy to dismiss the frenzy as just another tech fad, the latest toy for the Silicon Valley crowd to play with before moving on to the next shot shiny object. But here's the truth. AI isn't another fleeting trend. It's not a gimmick, it's not a meme. And it's not just about chasing the next viral app. AI is a foundational technology like electricity, the Internet, or the personal computer that's already rewiring how industries operate. Look at what's happening beneath the surface. AI is now powering logistics networks, crunching medical data, building safer cars, scanning financial markets, and transforming everything from agriculture to manufacturing. Corporations aren't investing billions for a quick dopamine hit. They're laying the groundwork for an entirely new economy. Companies like Nvidia, Microsoft, Google, and countless smallest players are racing to capture the infrastructure, the platforms, and the intelligence that will drive productivity for decades. And let's be clear, none of this happens without massive investment in AI infrastructure. Data centers, semiconductor fabs, energy grids, advanced networking. These are the backbones of the new digital economy. The demand for power, security, and scalable cloud storage is exploding. The companies providing the pipes, chips and protection for AI aren't just along for the ride. They are at the center of this tectonic shift. The gold rush isn't just in the apps, but in the picks and the shovels, the GPUs, the servers, the connectivity, and the cybersecurity that keeps it all running. You're missing the point if you think AI is just about chatbots and fun filters. This is about automating workflows, making more intelligent decisions, and unlocking value in places that we never thought possible. The winners in this space won't just sell software. They'll reshape what it means to compete in the global economy. Bottom line, AI is in a passing phase. It's the backbone of the next era of business and society. Investors who treat it as a toy will be left behind. Those who see it as the new industrial engine and understand the importance of the infrastructure will own the future. Okay. Joining us today is William Lee, research and analyst at Suro Capital, who focuses on identifying and supporting high growth opportunities in emerging technologies. Before Joining Suro in 2021, William worked at Deloitte and Washington D.C. leading predictive analytics projects and building financial models for nonprofit clients. He holds a master's in psychology focused on computational cognitive models and has a deep academic background exploring how humans and machines make decisions. Please join me in welcoming William to the show. William, it's a pleasure, an absolute pleasure to have you today. I look forward to this. Although I'm a little bit nervous about your psychology. Let's be real.

It's a pleasure to meet you, Kenny, and to be here.

Yeah, yeah. So listen, talk to me a little bit. Just give the, give the, the viewer just a little background on how your, your Deloitte training is led you into where you are today.

Yeah, I mean, so it's not just the Deloitte training, but I think academia as a background, as you focus on investments, they all sort of follow the same sort of idea, which is you have a thesis and from there you have to have supporting arguments that you have to test. And so as Surro Capital, when we go on investments, we look at sort of a very top down view. This is our thesis out there. This is where we think are sort of the core tenants. And then that's how we support our investment process.

Is sir rogue just about technology and AI?

No. So I can give you the genesis of Ciro, which is actually quite fascinating. So when you think about early 2000s, you had companies like Amazon and Google that going public between four and six years. What Ciro started or the reason why Ciro started was that once you get to about that 2010 timeline and you have the cloud and mobile revolution, companies are staying longer, private longer. So you have Uber or Airbnb for example. Those companies didn't go public until 10 plus years after the company started. And so you have this gulf now of about four to six years where companies are accruing massive amounts of value in the private markets. And you know, public market investors or retail investors aren't getting access to that. So that's the reason why Ciro started is to create a public vehicle. So, so that retail investors, public market investors can get access to these private.

Private investments.

Exactly. And so as it relates to AI investing, what Ciro is sort of seeing and what I think everyone sort of saw with, with mobile, is that there's all, there's this huge wave of technology that's coming out that is extremely valuable. And especially with the boom of ChatGPT at the late 2022, everybody knew that know this is a transformative technology. What people I think had a hard time, even on the investing landscape, had a hard time trying to figure out is, okay, there's all these companies that are starting, but which of them are here to stay. So Ciro took the view of looking at the companies from the infrastructure layer, companies that would be supporting AI enablement, generative AI enablement. And so we made three big bets within the space, which was in OpenAI, Coreweave and Vast Data.

Okay. And certainly OpenAI has now joined hands with Microsoft. Right. Core Weave came public, I don't know, a month and a half ago, two months ago, and has done nothing but shoot straight higher.

Yes.

Right. And continues. Actually, there's a bunch of banks this morning that came out and launched coverage and they all got buy recommendations with, you know, 200 plus price tags. I think it's trading 185 this morning. So they get. That's a significant. And that's up from where it came public. And so that's exciting.

And then your third one was, which is Vast data.

I don't think about vast data. So talk to us about that.

So when we look at the AI infrastructure layer, we think of sort of three key tenants. So you have the LLMs or the foundational models. So that's our investment in OpenAI is that there's all these applications that are getting built on top of this core technology. And so we wanted to make an investment there. The second part is compute. Right. So you need, similar to cloud, you needed compute to make any of these applications run. AI is very unique in that you need a special type of compute to be able to make that happen, which is these GPU clusters that are being built out on mass right now today. So that was our compute investment there. And then the third layer that we see, or the third pillar that we see is data, is that the way that our data infrastructure is set up right now was sort of set up to provide a good way to get consensus so you don't have duplicate data. That data doesn't run into one another, but it wasn't built for speed and for volume. And so what VAST is really doing, if you think of, just to bring a concrete example, if our current data infrastructure is like cross guard at an intersection, telling which cars to go and which people to cross or which people can cross, that's sort of a slower way of doing things. You can think of VAST as providing sort of a high speed highway so that the data is aware of where it should be going. There's signage, people kind of can do that on their own and the data can move to the right place at the volume and speed that AI is requiring of it.

Explain something to me because I'm just fascinated by this other kind of phenomenon that happens when they talk about AI hallucination and so explain to the listener, to the viewer what does that exactly mean and then how do you prevent that from happening? Sure.

So I think that's been a focus right now of a lot of enterprise AI agents and enterprise applications is this idea of hallucination. So the idea is that, you know, the AI is trying to make assumptions or the large language model is trying to make assumptions based on its training and a lot of different parameters that it has.

Right.

And so sometimes it, because it's not hard rules and hard facts about the world, it can sometimes make things up. And so what you makes things up, Right, right, exactly.

Or it'll conduct itself and that's the hallucination. Just so we understand when it makes it up. That's called the hallucination.

Exactly. It'll convince itself that it's correct. It's correct. Right. Because it's using a lot of generally soft or harder rules, but not like very hard line rules that like this is the truth about the world.

Right.

So now you see a lot of companies and where it's really exciting is that a lot of companies are trying to build, you know, these guardrails, ways to make enterprise applications more reliable. And so that's a huge investment area that I think will be burgeoning in the next couple of years.

Well, because it's interesting because to your point, you just said it makes itself believe, believe that what it says is true. And if somebody's using the AI who doesn't know. Right. Then the person reading it is going to say, okay, this must be true. Right, Right. So how does the person recognize that it's a hallucination or they don't.

So I think this is where you saw in what was, I think an improvement in sort of Google's searches at points in time where folks were using AI to get an initial start on Sort of a research query that they had and then they would subsidize it because they're like, hey, some of these things I want to double check or some things I want to even go deeper into. And so you see sort of like an acceleration of sort of both sides of people using AI to sort of help them get started or go deep and then also subsidizing that research as well.

Yeah. Because I think it's just fascinating. And I think, you know, sometimes people will use ChatGPT and then they'll cross check it against Grok or they'll cross check against Perplex or one of the other chat bot services. Right. To see whether or not it's getting the same response. And in fact it won't get the same response because all these different services are trained different ways. Am I correct?

Exactly.

And so there's a different database, actually. Or a different learning model.

Yeah. They have different parameters, different sort of settings that are associated with the model. Like you're exactly saying that would cause it to give different answers.

Different answers. Okay, hold on one second because we're going to take a break, but we'll be right back. All right, so let's just pick that up from where we left off. That the, that the, that the hallucinations are going to give or the models are going to give different answers if you, if you, if you check it across three different models. And so how do you know ultimately, as the user, where we are like, what's true and what isn't true if you don't understand or know what you're looking for?

Got it. Is the focus really trying to understand.

Yeah. A theory or, you know, say, say somebody's. I know. Trying to write it. Trying to look for, look to write a paper on history. Let's just, let's just break out history, right?

Yeah.

Although I'm not sure there'd be many hallucinations in history. Sure is history. Right?

Yeah.

So let's not talk about that. So let's talk about if they're trying to write a paper, if they're trying to understand the state of play in the AI industry. Right. Because I guess that's kind of a moving object. Right. How does it know? Like, I don't necessarily know. I'm not that sophisticated yet in understanding. So how am I going to know what's right and what's not right? Or I don't. I have to check. Do I have to actually check it against a human being and come to you and say, will help me out? Does this look. Right, right.

I think that there's folks that are, you know, looking and understanding that some of these responses, you know, aren't always accurate. But what you're seeing is a massive acceleration, at least as these AI models are getting better, that those errors are becoming less and less frequent.

Right.

And I think what is happening, and at least what we're seeing on the investment side is that a lot of money right now and focus is going into making these AI models or tools enterprise ready, which is less mistakes, trying to set up guardrails, hard rules so that it can give you accurate answers that you can trust and feel comfortable launching within your enterprise.

So do you have faith? Do you think that that's actually possible?

Yeah, I mean, if you think about ChatGPT back in 2022, it was a really exciting technology, there was a lot of challenges with it. But what you're seeing is that's as, you know, as bad as that technology will get. Right.

Right.

From that point on, things have only improved exponentially. And so you can, it's not hard to imagine that in a year, two years, three years from now, those types of errors that you're seeing will be. Yeah. Will basically not be non existent.

So tell me, let's talk about Turo for a minute and let's talk about what are the areas that Suro finds beyond LLMs that's going to get investor intention the second half of this year, next year, the year after.

Yeah, I think we still have a very keen eye on AI infrastructure. So that is still the compute and data layer, those two pillars. I think where we're starting to look beyond there now is, is moving up the stack. So now you're looking at development operations or some of these key companies that are now coming out and seeing massive adoption of their platform. So take for example Open Evidence. So they're basically a ChatGPT or AI assistant for medical and academic professionals. They have had and they've come out and along with, I believe Sequoia came out and said that they have about one fourth of physicians using open evidence today, which is massive adoption of these types of tools. And so where I think we're focusing a lot of our time is we're trying to find these tools either on the DevOps or infrastructure layer or even on the application layer where you're just seeing like it just makes sense, right, that people are adopting it and it's easy to integrate into your productivity work lines.

Well, I think that it's, I think it's amazing how, especially let's talk about the medical field. The medical field is adopting the use of AI in kind of analysis and interpretation of data, test results and kind of where they think this is going and how you might serve this patient.

Yes, right.

And I think that Brian, I know Brian Sauzi did a piece today with Microsoft's AI CEO and they were exactly talking about how it is so making a change.

Yes.

And that's, that's OpenAI, isn't it? Microsoft is OpenAI.

Yeah. Microsoft leverages OpenAI.

OpenAI. Right. So it's the OpenAI platform that, that they're using. But it's an amazing conversation. It was an amazing conversation that he was having with Mustafa and about, about where they're going with it and what it kind of looks like. But talk to me now about what, where do you think this is going in terms of the next wave of tools that are going to reshape the labor market?

Yeah, I mean, I think where it's happening first and as I mentioned with open evidence and there's other companies like Harvey, which raised around not too long.

Ago and a bridge and what's their specialty?

They're focused on. On law firms. Providing law firms. Yeah. Agents and assistance to law firms. And then you have a bridge which is focused on doing transcription for medical professionals, which are tasks that they spend way too much.

Yeah, way too much.

Right, right. But you see both of those companies within the last two months raise massive increases in valuation. So within six months raising sort of back to back rounds going from anywhere from about 3 billion to 5 billion.

Of those companies that Turo has an investment in.

We don't have investment in those companies. But I think it's very emblematic about what we're seeing on the private side, which is why I the way that Siro exists and that Ciro's existence is very fundamental. Right. So you have these companies raising rounds going from 3 billion valuation, 5 billion valuation. That's happening in the private markets.

Right.

So people don't have access to those, those types of companies. But where you're seeing a lot of this value accretion is in these private market names. And so that's where we turned a lot of our attention to is once those companies are starting to reach that massive adoption, that's when we get really excited.

Where do you think, where do you think human jobs are going to go in the future? Do you, do you think there's going to be just a change in the jobs that humans do? Or do you think there's going to be an elimination of jobs.

So, I mean, after many of these types of like, transformative technologies, you can think of industrial revolution or, and I'm sure many other people have given this example is people at those times thought that this is the end. Right. Machines have taken our jobs.

Right.

And so, I mean, people can speculate whether this time is different, but historically it proves out that the labor market is, is quite sort of resilient and adaptive to these types of, these types of productivity shifts. And so I think what you'll see is sort of the same type of shift in that there will be new types of jobs or maybe the types of jobs that are out there, things that we're not even thinking of today. So.

Right. So let's talk about where, where you think and maybe of a view on this is where's regulation going to go in terms of AI?

Yeah, I mean, so regulation, it seems like they've taken a little bit of a backseat to trying to figure out the best way to regulate AI. But I mean, it's definitely a moving target today. So it'll, it'll be interesting. There's sort of push and pull on what is sort of the best way to approach AI regulation because you don't want to stunt growth.

No, but there has to be, there have to be some guardrails on regulation, right?

No, absolutely. It seems like people are still trying to figure out what, I think all the pieces of AI technology and what are sort of the key components that really make up that recipe from the regulator, the regulator side. And I think from there they'll try to figure out the best way to regulate it.

But you know what the problem there is is it changes so quickly. The regulation can't keep up with it. By the time, you know, they're focus, and then suddenly it changes. They got to change. The whole, the whole conversation is going to shift.

Right, Exactly. And I think that's the argument on the other side of why would you set very hard and fast rules when this technology is still. I mean, ChatGPT came public, so came available to the public in, in 2020, late 2022. And so you're trying to now regulate this technology that's only been out into the public for only a few years now.

No, I hear. You know, it's funny because ChatGPT I think, was the very first one. But then there's, you know, perplexity and then there's Grok now on. By the way, I think that, I think that the Elon Musk's Grok Chatbot does a very good job. You know, I, I, I, I actually now kind of play them back and forth, chat GPT and grok to see to see who comes up with what answer and which one do I think actually answers the question better. And you know, I'm sure there are a bunch of others out there that I could be using or that I could be trying. But you know, you have to have I only have 24 hours in a day, right? In any event, listen, I want you to come back and talk about this again with me probably three or six months from now, because there's so much happening in the space that, you know, three, six, sometimes it's going to be completely different again. But I'm going to get at the end of every episode, I end my episode with a recipe. And so today I'm going to give you Pasta alla Norma, which is a classic tradition Sicilian dish. It was born in Catania, Italy, right, which is part of Sicily. Named after Vincenzo Bellini's opera Norma in 1831, it's said to have been created as a tribute to the composer of the opera's success. The dish emerged in the early 20th century with where it was first documented in a 1920s recipe book, though the exact origins are still being debated in this recipe, it features pasta, typically rigatoni and spaghetti. Now I use rigatoni. I think this demands a rigatoni, but you can use spaghetti. It's got a tomato sauce, fried eggplant, grated ricotta, salata cheese and fresh basil. The ingredients really reflect Sicily's agricultural abundance and culinary influences, blending the Mediterranean flavors with the Arab inspired elements like eggplant. Legend has it that a Catania chef, inspired by the opera's beauty, named the dish after it, claiming it was as perfect as Bellini's masterpiece. Today it is a beloved symbol of Sicilian cuisine, celebrated for its simplicity yet vibrant flavors. You can scan the QR code that's on the screen for the full recipe. You can thank me later. In the meantime, that's a wrap for today's Trader Talk, but the conversation continues. Subscribe on Apple Podcasts, Spotify, Amazon Music, or wherever you get your podcast. You got questions or topics you want me to cover? Email us@tradertalkahooinc.com because we're always listening. In the meantime, stay sharp, stay disciplined and stay in touch. And until next time, take good care.

This content was not intended to be financial advice and should not be used as a substitute for professional financial services.