Lead-Lag Live

Beyond the Hype: Derek Yan on the AI Build-Out, Private Access, and Where the Real Opportunity Lies

• Michael A. Gayed, CFA

Share episode

In this episode of Lead-Lag Live, I sit down with Derek Yan, CFA and Senior Investment Strategist at KraneShares, to cut through the noise surrounding the AI “bubble” narrative and unpack where the true opportunity still lies.

From infrastructure shortages and the rise of reasoning models to the challenge of accessing private AI leaders, Derek shares why the current cycle is still in its infancy and how investors can capture the next phase of growth.

In this episode:
– Why AI demand is outpacing computing capacity worldwide
– How reasoning models are reshaping the AI investment landscape
– Why the Magnificent 7 alone won’t capture the AI revolution
– The importance of balancing infrastructure vs application exposure
– How A-GIX connects investors to both public and private AI leaders

Lead-Lag Live brings you inside conversations with the financial thinkers who shape markets. Subscribe for interviews that go deeper than the noise.

#LeadLagLive #KraneShares  #AI #Markets #TechStocks #AGIX

Start your adventure with TableTalk Friday: A D&D Podcast at the link below or wherever you get your podcasts!
Youtube: https://youtube.com/playlist?list=PLgB6B-mAeWlPM9KzGJ2O4cU0-m5lO0lkr&si=W_-jLsiREjyAIgEs
Spotify: https://open.spotify.com/show/75YJ921WGQqUtwxRT71UQB?si=4R6kaAYOTtO2V

Support the show

SPEAKER_00:

So investors need to position the portfolio towards that AI ecosystem. And also, there's a dynamic allocation between infrastructure and application. That allocation is also very important for investors to really position their portfolio to really dynamically capture the best growth from the AI at each stage.

SPEAKER_01:

Right now, you can't turn on CNBC or Bloomberg without seeing headlines about artificial intelligence. And just as often warnings about an AI bubble. At the same time, trillions of dollars are being poured into infrastructure, data centers, and enterprise build-outs. For investors, the big question is whether this momentum is hype or the start of something truly transformative. My guest today is Derek Yan, CFA and senior investment strategist at Crane Shares. Derek works closely on AGIX, a strategy designed to capture opportunities across both public and private AI companies. Today, we're going to dig into how investors should be thinking about AI right now, where we are within the adoption cycle, and what might be coming next. Derek, welcome.

SPEAKER_00:

Thank you for having me.

SPEAKER_01:

So as I said, when you open CNBC or Bloomberg today, the phrase AI bubble is everywhere. How should investors interpret that? Where are we within the hike cycle? Or are we still sort of in the early innings of a more transformative trend?

SPEAKER_00:

Yeah, definitely. Like I think recently there's a lot of news on a concern of AI bubble, especially when you see like Nvidia invested in the OpenAI. OpenAI going to use the money to deploy to like Oracle, Oracle to buy more chips from Nvidia. So that circle just reminds me a lot of investors early-day internet that like Cisco providing capital to their clients. So but we think like for Nvidia, it's more like strategic investment because they found OpenAI's business pretty attractive. So that's way, I don't think like that that's really a bubble. In contrast, actually, if you listen to Jensen, if you listen to a lot of like stakeholders, CEOs in the AI industry right now, we have a shortage of computing. So just looking at a lot of if you use like OpenAI's ChatGPT, or if you're a soft engineer or software developers who use like a lot code like for like um your automation in the coding space, you realize there's come like a slowdown right now, like there's a shortage of um AI output because the demand is like tremendous. It's like you have, I don't know, it's like I think Jensen have this idea, their previous forecast is like it will be 10x, 100x demand for AI tokens output. But I think reality is like gonna be billions of billions X. So that's crazy. Just like imagine like how much demand is out there. So if like OpenAI anthropic can ramp up their capacity to provide their users the capacity to really compute, their revenue is gonna be double, triple. So that shortage, that demand is trickling a lot of investment into buying more computing power, then the computing power is going to buy more than video chips. So that's why you see the lot of financing demand from like model companies like OpenAI. So that makes me believe we're still in a white early stage of the AI build-out.

SPEAKER_01:

Derek, from an investment perspective, where do you think we are within the AI adoption and monetization cycles? Are we still in the infrastructure build-out phase or are we already moving into application phase?

SPEAKER_00:

Yeah, so that's a good question because I think like if you look at the past, but like when we have like the GPT moment, then the GPT model is basically pre-trained. So a lot of resources, like data, data laboring, um, then you need like a lot of chips and data center built out uh to train a model. Then you use that model. Um, so story is like that, and you have a lot, you need a lot of infrastructure investment uh to develop new models because new models can be better. However, that narrative really changed. Uh, I think earlier this year, in back in December last year, January this year, we have like OpenAIS 01 model, and you have the DeepSeq R1 model. So those new types of models called reasoning model, the cost like per uh disclosure by like DeepSeq, the cost of building a new model is much lower. So that's just like scared, I think, global investors, because like if that's the case, the models become like commodity. You don't need a lot of infrastructure to invest in the chips to build new models. The story totally changed. You have some like a panic there, then all of a sudden, like all the value migrate to the application layers. Oh, the software company, the SaaS company benefit. Um, because if the model is going to be so cheap going forward, the application is gonna benefit. But however, that narrative now shifts back again. Um, part of the reason is just like you do have that wave of efficiency gain for a software company because they deployed AI models. So their margin improved, there are a lot of workflow automated already. So those financial results are already being published, so people know that. So people don't think that's gonna go much higher because there's so much efficiency to gain. Um, there's I don't think that's a lot of ad right now. So people are looking for new avenue gain for those software, the application layer. But now with the new reasoning model, actually the inferencing demand, as I said, like people need new AI coding, people need the O1 to do more in the consumer-facing applications, the demand is like not enough. Just like the sorry, the capacity just not enough that we need more infrastructure and chips to actually just support that. Uh and like just look at a uni economy. Like, you think about like usually OpenAI or Anthropic, they quote um the clients, like in API, right? Like the million tokens per output. Like usually the price can be ranged like$12 or like$20. Um, then you have the like the cost is like currently it's lower. So you do have like unique economy um for each of the AI models. Um, so when you have that, just like if you can have chips, you have computing power, you can make more money. So I think like currently the infrastructure is back. And like that's where the investors are now focusing on the investment focus on AI.

SPEAKER_01:

Garrett, what would you say to investors who say, I already have exposure to the AI trade through the Magnificent 7 or the Nasdaq 100? What's the incremental case for adding a dedicated AI strategy like AGIX?

SPEAKER_00:

Yeah, I think like just similar to the internet era, right? Like you have um a long-term growth structure uh happening in AI, but the dynamic is just like more than I think seven names. Um when you look at even like Mac7 this year, like there's some dispersion. Um so and we are still the king. You have Meta um who's like um using AI to run ads, and those uh monetizations fast, and the they'll they have been performing well. However, then then there's some like lags. Um you have the Apple, then the Apple's strategy on AI, the talents uh is not ready. They don't have a good strategy. So you do have like some lagging performance, uh, even among Mac 7. Then it's even more obvious like among NASDAQ 100 because a lot of business within the NASDAQ 100 are going to be disrupted. Their business model is probably gonna lose the AI raise. Um, and their whole business can be just replaced by many AI native startups or other incumbents in the technology space. So if you think about that, like you do need to be very selective right now in the AI play going forward because it's not a Mac7, it's not NASDAQ, it's gonna be the AI winners. Like, who's gonna be where AI winners are gonna deliver those long-term structural growth? Um, so investors need to position the portfolio to really um towards that um AI ecosystem. And also, uh, I think like, as I said, you have to, there's a dynamic allocation between infrastructure and application. So that allocation is also very important for investors to really position um their portfolio uh to really dynamically capture the best growth from the AI at each stage.

SPEAKER_01:

Derek, just to get a little uh more deep into that, what are some of the biggest challenges in trying uh to get exposure to AI and how does AGIX help solve some of those issues?

SPEAKER_00:

Yeah, I think like there's three challenges. Um if you think about that, like one is just an allocation, I said. How do you really, even even though, even though like everybody knows, oh, AI is the play in the long term, you need adequate AI, but how can you do it? Then you have the AI hardware, you have AI infrastructure, you have you have AI applications. So, how do you really allocate the capital across each layer? That's uh very strategic, or sometimes can be tactical, dynamic. So you need a framework to really deploy the capital uh according to the current demand, right? So because you think about it, that's the the down upstream to downstream, then everybody's fighting for margin. So where's the bottleneck? So you really need to emphasize your where you need to really uh tilt your portfolio to that bottleneck. The second challenge is going forward, I think every company is gonna declare they're gonna be AI companies. Because everyone's gonna use AI. Um so a lot of business is gonna claim, oh, we're AI native, uh, we're gonna use AI. But uh in the end, they're just more than a height where there's their business model more likely to be disrupted by either the model companies, open AI is gonna destroy a lot of business, actually, then um many SaaSplay are actually gonna lead in many verticals. So we think you need to be, you need to think like AI native researchers, and you need insights from AI native researchers to talk to the AI researchers within each firm to gain that insight of each one's business model. Um that's the second challenge. Um, the third one is that most people's portfolio is only the public, right? So that'll give you a lot of like exposure to hyperscaler uh applications um like chips. But um one critical part of the ecosystem uh is missing, which is the foundational model companies, because most foundational model companies are private. Um however, that's where uh I think at this stage and going forward, uh most of the value is going to be created. So um many investors missing that value creation within their portfolio. So we as at creatures, we we identified those three challenges like two years ago. Um that's why we started this project to work with a bunch of AI native researchers um to create a portfolio to solve those three uh issues, I would think. So we we work with AI researchers. So those AI researchers they provide a score uh for each company within the universe that helps us navigate uh the dynamic of each company's business model, how the business model is ready for the AI disruption. And also the AI score is gonna be dynamic each quarter, so the weight towards each um category, uh hardware infrastructure application is gonna be uh dynamically reflected to position uh the best as the current stage of AI. And also because those AI native researchers, they are also venture investors, they have a close relationship with a lot of AI uh model companies. So we've been able actually to get access to a lot of private AI investment. Um, AGX, uh, on behalf of Clean Shires Trust, um uh we invested into the Anthropic um in March this year, uh in their CSE round. So this is a direct investment. We sit on the cap table. And also we invest in anthropic, sorry, the XI in July uh through their$10 for round. So we also sit on their cap table. So those um those relationships were critical, um, as well as those private investments, I think is critical to provide uh ecosystem that includes both uh the hardware, the infrastructure, applications, and models. So that gives you the investor the whole picture of AI.

SPEAKER_01:

I want to pivot uh just a little bit. Looking out, say three to five years, what's one thing about AI or the investment landscape that you think most investors are underestimating uh today?

SPEAKER_00:

That's a good question. I mean, like, first, I think like people keep underestimating the demand for AI. Um, just like all the think about even like Nvidia, um then their capex is probably like like shorting, providing the the to really accommodate the demand. All the hyperscalers, even they have like huge numbers in the capex. Like in the cloud era, like it turned out to be shortage. Now with all the infrastructure inferencing and reasoning demand, uh, it's shortage again. Um so we we're likely to see more like um underestimate of the demand for computing power and also the demand for AI. Um you have seen that like across Wall Street, I think starting in the internet era, um like most of the Wall Street like analysts, they they underestimate the demand for the Capax need. Um so that's number one. Um and number two, I think people now focus a lot of the kind of like digalization, um, because AI is really happening in the digital world. If you look at the industry, um, which industry is like now adopting AI penetration rate for the AI is highest among the technology sector, um, because technology sector, their workflow uh is all around code, around data. Um, so like if you think about like typical workload of software developers, product managers, those a lot of workflow can be automated using AI coding, AI agent. Um so that's that's a good opportunity right now. Um, but if you look down, um, a lot of industries, their penetration is quite low. Uh, think about like manufacturing, think about like healthcare, a lot of service sector. Um, I think those are potential opportunities for AI going forward in more longer term. Because think about like the AI can be two things. One is the uh digital AI, but then now we increasingly see more chance from the physical AI. Um now you have like Elon Musk trying to develop like R2D2 for everybody. So like the humanoid robotics, like we have like a humanoid robotic phone as well, uh, called uh KOID coid. That is really investing in the ecosystem to build out those like R2D2 components. Um that like we bring one of the humanoids actually to green the belt NASDAQ. So once you see the performance, the accuration system for the humanoid, it can run, it can move, it can wave, it can shake hands. Like human now uh is much better compared to like years ago, where you probably see this like footage from like Boston dynamic. Uh, but if you look at like in the real real life, the performance of the humanoid robotics right now, today, it just will be shocking. So I think that just like give everybody like optimistic um uh outlook uh on the physical AI, because now when you have AI models can incorporate like action data, our environment data, um, image data with the vision technologies, the sensor technologies, you can actually create like a whole physical world AI. Um the factories, like if you look at the latest factories, they have like digital twins, right? So they create a digital factory. So like we can do that like for every everything going forward. Like even when our human body can have a digital twin, so all the healthcare robot can work, right? So that's that's something I think like is really underestimated or missed position in everybody's portfolio. Um, so um I would think like uh as uh as cranchers, um, that's why we launched uh KOI decoy uh as a player to really tap into that early opportunity.

SPEAKER_01:

It was epic uh to see the footage of that humanoid at the New York Stock Exchange. Derek, just to before we wrap up, for anyone who wants to learn more about AGIX or to connect with you or the team, where's the best place for them to go?

SPEAKER_00:

Yeah, they can um uh craneships website, we actually publish a lot of research on AI and we have a white paper on the AGIX of FAQ. So people can go to Cranchis.com slash AGIX. Um where anyone wants to know COID, they can go to Craneships.com slash COID. Um, yeah, just like that's the go-to place.

SPEAKER_01:

Fantastic. Well, Derek, thank you so much uh for joining me again, and thanks to everyone for watching. Be sure to like, share, and subscribe for more episodes of Lead Live Lives. See you next time.

People on this episode

Head Shot

Michael A. Gayed, CFA

Host