AI Themes to Watch

Connor Round: Hello, everyone and welcome to Public Live. I'm your host Connor Round and I'm so happy to be here with you today to discuss the key themes in AI and the industries that it can bring innovation and growth to. I'm thrilled to welcome my guest today, Ivana Delevska, Founder and Chief Investment Officer at Spear Invest.

Ivana Delevska: Thanks for having me, Connor. 

Connor Round: So just to start off, you know, we're here to talk about artificial intelligence. It's an opportunity that has use cases beyond the human imagination. So just to start off when considering business applications of AI, could you help investors understand what industries and sectors you see greatly benefit from the innovations in the space?

Ivana Delevska: Yes, sounds great. Connor so many people under appreciate the fact that AI today is already a $100 billion industry and it's still in the very early innings. The main applications that we see it used in today would be for consumer technologies. So it would be for like recommenders in cases like TikTok or e-commerce where you're getting recommended products to buy. So that's one of the largest applications that we see today. But looking forward, we think AI will be transformational for many industries, traditionally industrials, automotive, life sciences. We think it's really going to have a transformative impact for the broader industrial complex.

Connor: Very interesting. And could you just break down the AI capabilities that you see within the social media space? 

Ivana Delevska: Well, yeah, so it's interesting social media. Previously, all of the innovation or all of the use cases in AI were based on this recommender model. So basically, companies like TikTok would in-house train their their models to be able to tell based on few data points on videos that you watch, they're going to be able to form a whole profile about you, and then be able to, to recommend you other videos, and they can be really addictive as most people, most people have seen. 

So that's kind of the main use case today. Now, the new thing in AI is this transformer model, which was the concept or the paper was written and introduced in 2016. But you've really seen these models just getting trained, these large language models. So Chat GPT, is one application of this large language model, it's based on a model called GPT 3 that was introduced in 2021. So a lot of people are excited about the applications of Chat GPT today, but I will highlight that this is just one kind of flavor of the opportunity that we see going forward.

Connor: Great, I'm actually happy you brought up chat GPT that was just gonna go right into that. Apart from it kind of just expanding and, you know, breaking out into the scene very quickly. How do you see other companies reacting in the space? Do you see Google coming out with a new search engine capability? I'd love to hear your opinion on that. 

Ivana Delevska: Yeah, so we see Microsoft and Google, as potentially being leaders on the language side, Microsoft has maybe a head start, they've invested quite a bit here. And we've seen quite a lot from them. They're behind the GPT 3 language model, but they also recently announced an even newer model with even more parameters, 530 billion parameters, so we're gonna see a lot more from them. 

Google has been pretty quiet. They were maybe the leader several years ago. So we are waiting to hear what's going to come out of them. Investors believe that they have something, but it's kind of TBD. So as of right now, those are the two that are going to be leaders in the lead on the language model side. But it's very important for investors to realize that when you invest, you don't necessarily just want the direct beneficiary over trend, there is going to be a lot of investment in the infrastructure beyond behind like, behind AI. So for Microsoft, it's not just on the Bing side that they can introduce Chat GPT and kind of help jumpstart their business, but they're really going to benefit on the cloud side as well. So from an investment perspective, right now, we're finding better risk rewards in investing in the infrastructure. So it would be semiconductors, cloud infrastructure, and cybersecurity as well. So yeah, so basically, the point here is beware of value traps, right? You don't want to invest in something that will get competed away.

Connor: Yeah, definitely. And just to keep on Chat GPT, a lot of the hype has brought excitement to firms. And it kind of seems to be a similar trend to that of the crypto boom in 2021, with BuzzFeed announcing Chat GPT similar implications. Just wanted to know, do you see this as maybe a trend that will die down? Or is this something that we can expect to last?

Ivana Delevska: Well ChatGPT specifically, what you will need to see is people building things that are going to be useful on it. So specifically, for us, we're looking for maybe broader opportunities than that. But AI is a trend that is definitely not going to die down. As I mentioned before, it's already a $100 billion dollar industry. So whether Chat GPT takes off or not, it’s not really going to have..it will have some implications on AI but it's not really going to have broad implications for AI, if that makes sense.

Connor: Definitely and as you said, it is a $100 billion industry already. But in order to rev up innovation in this space further, much hardware and infrastructure will be required, such as the need for, you know, the much reliance semiconductor industry. 

What types of semiconductors, so we could just help investors understand, are required for building out AI, and how much of the industry is dedicated to its face? 

Ivana Delevska: Yes, so, so thanks for asking that question, Connor. So basically, in order to train these models, you need GPUs. And that's why companies like Nvidia produce these today. And basically GPUs allow computers to do parallel operations and the reason why people are all of a sudden so excited about AI versus say a year or two ago is that prior to the transformer model, and prior to this language, first language models, trained AI will just be able to do prescriptive tasks. So, if you give it a bunch of images, it can learn what you're showing it, right? So it wouldn't necessarily be able to put things into context or draw its own conclusions. So in order to be able to put things into context and draw its own conclusions, the model sizes have increased exponentially. So maybe five years ago, in 2018, the model sizes would have been less than 100 million parameters. Today, these models that are trained and deployed are over over 500 billion parameters. So the amount of GPUs that are required for this is exponential.

Connor: Great, and what players in the space do you see benefiting from increased demand for AI related chips such as GPUs?

Ivana Delevska: So in video is gonna be a big beneficiary full disclosure, we all not only video stock, and we disclose all of our holdings on our website, but then you may be able to see the reason why Nvidia is such a beneficiary is because they came up with this CUDA architecture that is free for people to use. So if you want to build something on AI, all you need to do is buy the NVIDIA hardware, and then you can use these frameworks for free. Other companies are not there yet. So you could still buy GPUs from AMD. But then what do you do with them? But over time, as the industry develops, there is likely going to be other solutions, and maybe some of these other companies that are developing GPUs will benefit. 

Another area we really like that is not doing very well right now, is cloud infrastructure. So a lot of the cloud names are getting hit pretty hard, because of enterprise spending cuts. So this is a sector that hasn't really done that well in this recent bounce, this year. Nvidia is already up quite a bit here today, but the cloud cloud names have really not performed in line, so we like that. And then as you deploy AI, you're going to have a lot more vulnerabilities so you're going to need newer cybersecurity solutions. So this is another area where we find a lot of opportunities – basically how do you secure? How do you secure these AI workloads? And how do you secure eight edge computing?

Connor: Great, so a lot of industries that are going to see some upside there and those segments, but the industry has struggled with keeping up with shift demand through supply chain constraints, severely dealing with shortages throughout the years of 2021 and 2022. If AI demand gears up in the near future, will semiconductor firms be able to manage their supply chains to keep up with demand trends?

Ivana Delevska: So we see a lot of supply actually getting added over the next two to three years. So we think there is going to be plenty of supply of chips available. And just to explain how this works, basically companies like Nvidia and AMD, they don't manufacture the chips, but they just design them and then foundries like TSMC will do the actual manufacturing. So we're seeing - this is a pretty big trend actually that's supporting even broader economic activity in the U.S. are these large CapEx investments by the foundries or the large chip manufacturers like TSMC. Intel as well has expansion plans. So there's going to be plenty of supply, we believe to meet the demand.

Connor: Great And just lastly, could you just break down the traditional segments of the semiconductor space? What spaces or segments do you see slowing down? And are there any other ones that you see an opportunity, and I know that you named a few, but if you just break them down for investors to understand a little bit more, I think that'd be fantastic.

Ivana Delevska: Yeah, absolutely. So basically the big separation is between consumer and data center. So, if you look at, for example, TSMC, revenues would be like half consumer, like a very high level, half consumer half data center. So consumer would be used in PCs, smartphones, and then on the data center side, this would be like cloud, cloud vendors, and in any sort of enterprise computing. And then within the data center segment, we're seeing slowdown across the board. So the consumer got pretty hard hit at the beginning of last year, PC demand has been significantly down but the way semis work is usually there is a lot of inventory in the system. So what the manufacturers or designers see is a lot worse than what in demand looks like. 

So as inventory stabilizes, TSMC said that they're expecting it to stabilize at the beginning of this year, and to see pick up in the second half of next year. And then similarly, on the data center side, that segment just started seeing some weakness, but we believe that the high performance computing, which is what we talked about with Nvidia, like the chips that are required for AI, we expect those will hold up better.

Connor Round: Great. Well, everyone, that's Ivana Delevska, Founder and Chief Investment Officer at Spear invest. Thanks for tuning in, folks. We'll see you next time.

Ivana Delevska: Thanks, Connor.

Connor Round: Thank you.