Animal Spirits Podcast

Derek Yan discusses Anthropic's S-1, KraneShares' private investment, and AI infrastructure shifts.

27 min Listen to the original
AI InfrastructurePrivate EquityETF StrategyAnthropicKraneSharesAutonomous Driving

Summary

Derek Yan of KraneShares joins the Animal Spirits Podcast to discuss the company's investment in Anthropic via the AGX ETF. He highlights data from Anthropic's S-1 filing, noting low Fortune 500 penetration and improved compute efficiency. Yan argues that AI infrastructure risks have shifted from overbuilding to underbuilding due to hardware and energy constraints. He also explains the AGX ETF's strategy of targeting the broader AI ecosystem with specific allocation caps.

Key takeaways

  1. 1
    To me, the risk of overbuild is now shifting to underbuild

    Unidentified speaker (speaker_3) · 8:56

Episode sections

  1. 0:00 – 0:44

    Intro and Sponsor

    Episode introduction with sponsor read for KraneShares AGIX ETF and standard disclaimers.

  2. 0:44 – 1:19

    Guest Introduction and Anthropic S-1

    Hosts welcome Derek Yan from KraneShares to discuss the Anthropic S-1 filing and KraneShares' investment.

  3. 1:19 – 3:13

    KraneShares AI Portfolio and Private Investments

    Derek explains KraneShares' ownership of Anthropic and other private AI companies, highlighting physical AI as the next growth phase.

  4. 3:13 – 5:45

    Anthropic Investment Details

    Derek confirms KraneShares' ~$17M investment in Anthropic via the AGX ETF from the Series E round at a $61.5B valuation.

  5. 5:45 – 8:06

    Anthropic S-1 Analysis: Revenue and Efficiency

    Derek highlights low Fortune 500 penetration (0.1% of expenses) and strong efficiency gains (5x compute ROI improvement).

  6. 8:06 – 10:05

    AI Infrastructure: Overbuild vs Underbuild

    Derek argues the risk has shifted from overbuilding to underbuilding due to hardware, energy, and capital constraints.

  7. 10:05 – 11:27

    Capital Costs and Market Dynamics

    Discussion on the cost of capital for AI infrastructure, noting hyperscalers have lower financing costs.

  8. 11:27 – 26:53

    AGX ETF Strategy vs Broad Indices

    Derek explains AGX's focus on the broader AI ecosystem and its 40% cap on single layers to manage risk.

Assets discussed

AssetKindStanceContext
Anthropicprivate companyneutralWelcome to Animal Spirits with Michael and Ben. Well, it's here. We saw the preliminary S1 from Enthrop. Anthropic.
SpaceXprivate companyneutralAnthropy definitely has now become kind of like where known one. We actually invest in XAI as well in July 2005 and December 2005. Those shares become part of the SpaceX and IPO'd. So we already had one exit out of the our private investment. Um then we have actually now more investment in the physical
NVIDIANVDAequityneutralYeah, exactly. Like when we launched Ajax, that's kind of like a pop-up top-up online. on mine for SP NASDAQ. That's so concentrated in Nvidia and hyperscalers.
TeslaTSLAequityneutralBut when you have intelligence improving, not not just from Tesla, but so many other players look at like physical intelligence, skid AI,
S&P 500indexneutralThis might be outside the realm of like what you typically talk about, but the SP 500, for example, is obviously heavily skewed towards this theme already.

“Stance” is how the speakers talked about the asset, not a recommendation.

Claims worth verifying

KraneShares invested in Anthropic's Series E round in February 2025 at a valuation of $61.5 billion.

Unidentified speaker (speaker_3) ·3:41

How to verify: Check KraneShares AGX ETF holdings disclosure or SEC Form N-PORT for private asset holdings; verify Series E valuation via Crunchbase or PitchBook. For recent events, verify 2026-08-16 through 2026-11-16. Use another period only when the speaker explicitly names it.

Primary source

Anthropic's clients include more than half of the Fortune 500, spending less than 0.1% of their total expenses on Anthropic.

Unidentified speaker (speaker_2) ·6:17

How to verify: Verify via Anthropic's S-1 filing on SEC EDGAR, specifically the 'Risk Factors' or 'Management's Discussion and Analysis' sections regarding customer concentration and revenue drivers. For recent events, verify 2026-08-16 through 2026-11-16. Use another period only when the speaker explicitly names it.

Primary source

Anthropic achieved GAAP profitability in Q2.

Unidentified speaker (UNKNOWN) ·7:19

How to verify: Check Anthropic's S-1 filing on SEC EDGAR for audited financial statements showing net income for Q2. For recent events, verify 2026-08-16 through 2026-11-16. Use another period only when the speaker explicitly names it.

Primary source

Anthropic's compute efficiency increased 5x over a year, generating $2 per $1 spent on compute.

Unidentified speaker (UNKNOWN) ·7:19

How to verify: Verify via Anthropic's S-1 filing on SEC EDGAR, looking for metrics on compute utilization and revenue per compute dollar. For recent events, verify 2026-08-16 through 2026-11-16. Use another period only when the speaker explicitly names it.

Primary source

KraneShares AGX ETF has a 40% cap on allocation to any single layer (hardware, infrastructure, applications).

Unidentified speaker (speaker_3) ·13:20

How to verify: Check KraneShares AGX ETF prospectus or fact sheet on KraneShares website or SEC EDGAR for investment restrictions and portfolio composition rules. For recent events, verify 2026-08-16 through 2026-11-16. Use another period only when the speaker explicitly names it.

Primary source

SEC guidance limits illiquid assets in ETFs to 15%, with KraneShares targeting 10%.

Unidentified speaker (UNKNOWN) ·20:34

How to verify: Verify SEC Rule 22a-4 or relevant ETF regulations regarding illiquid assets; check KraneShares AGX prospectus for illiquid asset limits. For recent events, verify 2026-08-16 through 2026-11-16. Use another period only when the speaker explicitly names it.

Primary source

xAI shares became part of the SpaceX IPO.

Unidentified speaker (speaker_3) ·4:23

How to verify: Verify via SpaceX IPO documentation or press releases regarding the inclusion of xAI shares in the offering. For recent events, verify 2026-08-16 through 2026-11-16. Use another period only when the speaker explicitly names it.

Primary source

The S&P 500 is approximately 40% concentrated in the top 10 names.

Unidentified speaker (speaker_0) ·11:27

How to verify: Calculate using S&P Dow Jones Indices data for the S&P 500 index composition as of the recording date. For recent events, verify 2026-08-16 through 2026-11-16. Use another period only when the speaker explicitly names it.

Nvidia and hyperscalers are heavily concentrated in the Nasdaq 100.

Unidentified speaker (speaker_3) ·11:54

How to verify: Check Nasdaq 100 index composition data for weightings of Nvidia and major cloud providers. For recent events, verify 2026-08-16 through 2026-11-16. Use another period only when the speaker explicitly names it.

Primary source

Do your own research

  • Verify Anthropic's S-1 filing for official revenue, expense, and profitability figures.
  • Check KraneShares AGX ETF holdings and illiquid asset disclosures in SEC filings.
  • Confirm Anthropic's Series E valuation and KraneShares' ownership percentage via public cap table data or press releases.
  • Review Anthropic's Q2 GAAP financial statements for compute cost and efficiency metrics.
  • Analyze AI infrastructure capex trends from hyperscaler earnings reports to assess overbuild vs underbuild risks.
  • Evaluate the AGX ETF's 'AI score' methodology and rebalancing rules in the fund prospectus.
  • Compare robotaxis unit economics against other autonomous driving applications using industry reports.
Full transcriptexpand
0:00
Unidentified speaker (speaker_0): Today's Animal Spirits Talk Your Book is brought to you by Crane Shares. Go to Craneshares.com slash AGIX to check out the Crane Shares public private AI and technology ETF. Ticker AGIX. It's Craneshares.com slash AGIX to learn
0:13
Unidentified speaker (UNKNOWN): more.
0:14
Unidentified speaker (speaker_1): Welcome to Animal Spirits, a show about markets, life, and investing. Join Michael Batnik and Ben Carlson as they talk about what they're reading, writing, and watching. All opinions expressed by Michael and Ben are solely their own opinions. opinion and do not reflect the opinion of Ridholtz Wealth Management. This podcast is for informational purposes only and should not be relied upon for any
0:34
Unidentified speaker (speaker_1): investment decisions. Clients of Rid Holtz Wealth Management may maintain positions in the securities discussed in this podcast.
0:44
Unidentified speaker (speaker_2): Welcome to Animal Spirits with Michael and Ben. Well, it's here. We saw the preliminary S1 from Enthrop. Anthropic.
0:51
Unidentified speaker (UNKNOWN): The
0:51
Unidentified speaker (speaker_2): 300-pound gorilla came down from the mountain, showed us its silver back, and said, Here I am.
0:59
Unidentified speaker (UNKNOWN): Good
0:59
Unidentified speaker (speaker_0): analogy. Better than expected, I think. So on today's show, we brought uh Derek Jan back on. Derek is from a street senior investment strategist at Crane Shares. And Crane Shares actually invested in Anthropic, I guess at a much lower valuation. So they they have an ETF that invests in public and private assets. and they have a handful of private investment companies in the AI infrastructure, and
1:19
Unidentified speaker (speaker_0): Anthropic is one of
1:20
Unidentified speaker (UNKNOWN): them.
1:20
Unidentified speaker (speaker_2): This is a little teaser, a little preview for the show. If you're wondering why Ben and I just laugh so hard, it's because we're we're recording the intro after we recorded the episode. And in the episode, I stumble over. Well, you'll see.
1:31
Unidentified speaker (speaker_0): Michael messed up his elephant and gorilla analogies. So anyway, yeah, great timing to have Derek on because they have been an investor in Anthropic. And it it's pretty wild that you can now have these private companies within an ETF structure, right? The stuff that they can do in ETFs now is is kind of insane. We talked a little bit about how you build an AI portfolio. And I think the interesting thing he said to
1:51
Unidentified speaker (speaker_0): me was that they have some limitations. Like they can't let just one part of the AI build out About part take takeover, over. Like they have they have ceilings on these things. So they're not gonna let one piece become 80% of the portfolio. I think having that sort of risk management probably makes a lot of sense now between the hardware and software and the frontier models and all this stuff. So last time we had Derek on, we talked about,
2:09
Unidentified speaker (UNKNOWN): I
2:09
Unidentified speaker (speaker_0): think we talked about robotics the whole time and how that's the next step. The future's gonna be very weird.
2:14
Unidentified speaker (speaker_2): That's
2:14
Unidentified speaker (speaker_0): all I can say. It's gonna be cool, it's gonna be weird, it's gonna be messy getting there probably, but it's gonna be interesting. Anyway, we talked talk about AI and robots and all and anthropic S1 and all this stuff with Derek from Crane Shares. So here's our talk with Derek.
2:30
Unidentified speaker (speaker_2): Derek, welcome back to the show.
2:32
Unidentified speaker (speaker_3): Thank you for having me. This
2:33
Unidentified speaker (speaker_2): is very commonly discussion. We are recording this on Tuesday, October 27th. We got Anthropic's preliminary S1 yesterday, and we're talking today about, you know, crane shares. You guys have great tickers. I always joke with you guys about K-Web, K Web, Queb. I'm looking at this ticker, I'm like, huh, AGIX, Agex, AGEX. AGEX, what is that? And then I realize it's
2:53
Unidentified speaker (speaker_2): AGIX, so you can't fool me.
2:56
Unidentified speaker (UNKNOWN): We
2:56
Unidentified speaker (speaker_3): had
2:57
Unidentified speaker (UNKNOWN): a
2:57
Unidentified speaker (speaker_3): lot of debate on the AGI, like
3:00
Unidentified speaker (UNKNOWN): some people
3:01
Unidentified speaker (speaker_3): really get AGI. Yeah, okay. Yeah.
3:03
Unidentified speaker (speaker_2): It's a it's a good ticker. All right. So let's let's start with the uh the gorilla in the room. Or the elephant in the room, I should say. 300-pound gorilla, elephant in the room, right?
3:13
Unidentified speaker (speaker_2): All right. How do you guys own an Enthropia? anthropics? So you guys own about $17 million worth, give or take again. This is Tuesday, October 27th. Where are the shirts coming from?
3:21
Unidentified speaker (speaker_3): We launched uh AGX in 2024. We come like thinking about like, yeah, we we gotta own the AI ecosystem, but what's the missing part? Is the model company. So we got like invest in the private through our partnership. We come like know anthropics like finance. financing. So we did participate in
3:41
Unidentified speaker (speaker_3): their series E round of financing in February 2005. AGX is actually on the cap table of the Anthropic. We got in at a valuable $61.5 billion. Today they're going to PO.
3:52
Unidentified speaker (speaker_2): Wow. Home run.
3:53
Unidentified speaker (speaker_3): Yeah, it's very good round for the AGX shareholders.
3:56
Unidentified speaker (speaker_0): I want to talk a little bit about the Anthropics Anthrobics S1, but any other private companies you guys own in the fund or not? Or is Anthropocopic Anthropic the only
4:02
Unidentified speaker (UNKNOWN): one?
4:03
Unidentified speaker (speaker_3): Anthropy definitely has now become kind of like where known one. We actually invest in XAI as well in July 2005 and December 2005. Those shares become part of the SpaceX and IPO'd. So we already had one exit out of the our private investment. Um then we have actually now more investment in the physical
4:23
Unidentified speaker (speaker_3): AI names. Uh we invest in autonomous a tons-driving driving company called New
4:26
Unidentified speaker (UNKNOWN): York.
4:26
Unidentified speaker (speaker_3): Neuro, a humanoid company called Aptronic, and a physical AI model company here in New York called General Intuition. So we feel like physical AI is the next phase of AI growth. That's where our private investment focused recently.
4:42
Unidentified speaker (speaker_2): All right. So we're going to talk more about humanoids in the show. Elon had a crazy quote. I think it was a the test earnings call. I can't remember which one it was. People that have helped held this ETF since you bought Enthropic have done quite well on that investment. They're trying to go public, you know, between one and a half to two trillion. We'll see where the market shakes out. What is the plan? I assume that once this is public,
5:03
Unidentified speaker (speaker_2): you're not gonna right. This has got to be part of the portfolio for the duration.
5:07
Unidentified speaker (speaker_3): Exactly. Similar to our SpaceX, For
5:10
Unidentified speaker (speaker_3): we're gonna continue to hold it as one of the holdings, also subject to lockup because. our private holding for anthropic is for sure the most important infrastructure within the ecosystem. So we're gonna continue to hold anthropic.
5:22
Unidentified speaker (UNKNOWN): So
5:23
Unidentified speaker (speaker_0): Michael and I were talking about the Anthropic S1 this morning on Animal Spirits, and we're not necessarily forensic accountants by any means. But what do you think stands out and what's important for people to know in the numbers there? Because it's a really long document. Obviously, a lot of people just put the document into AI and ask for the bullet points. But what do you what do you think people can pull out of there that maybe is kind of
5:42
Unidentified speaker (speaker_0): misunderstood or people don't really get about the business?
5:45
Unidentified speaker (speaker_3): I think the two key points like as my takeaway. Um, one is about the growth, right? Like people thought like AI is slowing down. But one thing I noticed is their revenue, they have clients that uh are more than half of the fortune. 500. But those Fortune
6:05
Unidentified speaker (speaker_3): 500 spent part of the total expense, less than 0.1% into Anthropic. So the penetration is there by the dollar spend is was way lower than people thought.
6:16
Unidentified speaker (UNKNOWN): Wait,
6:17
Unidentified speaker (speaker_2): hold on, Derek, Derek, could you just could you repeat that data point? Because that is that is a big one I haven't heard before.
6:23
Unidentified speaker (speaker_3): Yeah, so think about like the clients of anthropic right? Everyone knows their enterprise player. So more than half of SPO Fortune 500 are their clients. But those clients only spend about 0.1%
6:41
Unidentified speaker (speaker_3): to Anthropic as part of their total expense. I would say, like, if you look at their top 10 Fortune 500 clients, that's 1%. So if think about like the penetration gonna increase and further spending increase, you can easily have like 10x or 20x more spending from those companies.
7:01
Unidentified speaker (speaker_3): So I would say that's something to me. The AI spending is like beginning. With that beginning, you can achieve like 60, 70, or 100 billion AR. ARR, That's impressive. That's one takeaway. I was like sure. shocked. The second takeaway I would say is the efficiency gain.
7:19
Unidentified speaker (UNKNOWN): Right.
7:19
Unidentified speaker (speaker_3): A lot of people said, okay, you cannot make money from AI, you can lose money. But using the gap standard, they achieve profit in Q2. And look at like the efficiency of band on the compute. Actually, that efficiency has have increased 5x over a year. So, which means they can probably
7:39
Unidentified speaker (speaker_3): generate like $2 per $1 span on compute. So the profitability is not an issue if this trend continue.
7:46
Unidentified speaker (speaker_0): Derek, just thinking through Michael and I've been talking a lot about the risk factors here. I don't think a lot of people are talking about, well, tech people are talking about the upside, but like, what do you think is the most misunderstood point of what's going on in AI in terms of the people who are bearish? Because there's a lot of people who say, hey, listen, it's circular financing. it's all the hyperscalers going all in, and it's just trading money
8:06
Unidentified speaker (speaker_0): around. And uh what do you think about that and how it goes to sort of the next level?
8:11
Unidentified speaker (speaker_3): Yeah, we have a lot of pullback, right? Like look like the history of last three years. We have pullback in April 2005, we have November 2025, just recently, July, just last few weeks, had a lot of volatility. The previous concern was really like are we overbuilt? Are we like the dot-com bubble? But like
8:31
Unidentified speaker (speaker_3): right now, the more concern is uh like are we underbuilt?
8:35
Unidentified speaker (UNKNOWN): Um,
8:36
Unidentified speaker (speaker_3): because demand is so large. But you have so much constraint, right? You have one constraint from the physical hardware side, the HBN, TSMC, you have constraint from the energy side, you have politics, where you have local governments pushing back. Now you have capital constraint. You have the private credit, you have all those like
8:56
Unidentified speaker (speaker_3): high-ear bond markets, like you need to provide enough capital to the AI data center build-up. To me, the risk of overbuild is now shifting to underbuild, which means
9:08
Unidentified speaker (UNKNOWN): the
9:09
Unidentified speaker (speaker_3): value can migrate from the hardware to infrastructure, those who already have capacity. their margin can improve, right? Like GPU rental price is going higher. It might slow down the overall progress as those constraints kicking. But the long-term picture is, I think, a total adjustable market,
9:29
Unidentified speaker (speaker_3): like look at AI, the $40 trillion knowledge workers, all those like penetration at very low point. Is like I think like investors have to think about it long-term, long term. where, where is the money going to be made? Who's gonna be player making that margin? And where we are today. You wouldn't worry so much about short term water today.
9:46
Unidentified speaker (speaker_2): There's two things that you hear the bears say. Number one, I forget who said this quote the data centers are going to be turned into pickleball. Well, it's part of the same thing, it's gonna be turned into pickleball courts if they can't fund, if they can't finish the build out. So it sounds like you're not buying that on the demand side, but what about the capital, the cost of capital side?
10:05
Unidentified speaker (speaker_3): Yeah, that's the that's a risk. I mean, for depends on which player you're talking about, right? Like new clouds, they have a higher now financing cost because investors just like they have demanding higher returns, the risk going higher, like the junk belt market. So it depends on the players. For hyperscalers, they have very low
10:24
Unidentified speaker (speaker_3): financing cost. For like you Elon Musk, um, I think of like a different story.
10:29
Unidentified speaker (speaker_2): But even Meta has to issue equity.
10:31
Unidentified speaker (speaker_3): Yes.
10:31
Unidentified speaker (speaker_2): To pay for this.
10:32
Unidentified speaker (speaker_3): Yes.
10:33
Unidentified speaker (speaker_2): So not not that cheap.
10:35
Unidentified speaker (speaker_3): Exactly. So I think that's the capital market need
10:38
Unidentified speaker (UNKNOWN): to
10:38
Unidentified speaker (speaker_3): figure this out. When you have that demand and when you have that return, the question is like who's gonna provide capital? That becomes Become now a constraint. Then the return is gonna be much higher given that higher margin and higher demand.
10:51
Unidentified speaker (UNKNOWN): So
10:51
Unidentified speaker (speaker_3): I think there will be like a trade-off here. There's a tension that investors have a higher expectation, higher requirement, but opportunity becoming more attractive. So this is gonna be like a attack along the way.
11:07
Unidentified speaker (UNKNOWN): A
11:07
Unidentified speaker (speaker_0): lot of people are talking how AI is just this one big trade. It's part of the economy now, it's part of all the markets. I mean, the SP is 40% in the top 10 name. names. The NASDAQ 100 is probably 50 some percent in the top 10 names. Beyond owning some shares in the private companies, how do you distinguish this fund from like the Nasdaq 100? Like, do you own more global stocks as well, like the
11:27
Unidentified speaker (speaker_0): SK high inexes
11:28
Unidentified speaker (UNKNOWN): of
11:28
Unidentified speaker (speaker_0): the world and such? Like how do you distinguish from the Nasdaq that is already very heavy on hyperscalers?
11:34
Unidentified speaker (speaker_3): Yeah, exactly. Like when we launched Ajax, that's kind of like a pop-up top-up online. on mine for SP NASDAQ. That's so concentrated in Nvidia and hyperscalers. Um, but we feel like the Vatican migrate today is more mostly just like certain hardware or infrastructure, but it's broaden up
11:54
Unidentified speaker (speaker_3): to like say data monitoring, say the A model infrastructure, say the hyperscure hyperscurity cybersecurity, then to application, application can today be more agent-centric, but then in the future can be more fiscal AI-centric. So the value migration is something we think investors,
12:13
Unidentified speaker (speaker_3): they have to re either think about themselves and do the work to rebalance themselves, or you can use ETF like AGX. We have a systematic rule that we leverage AI score that help investors. navigating those dynamics along the time. You capture the broad ecosystem so it's not concentrating a few names. And also
12:33
Unidentified speaker (speaker_3): it's gonna migrate that value along the time, besides the private allocation.
12:38
Unidentified speaker (speaker_0): What are you looking for in that score, Derek? We
12:40
Unidentified speaker (speaker_3): look at like several things. One is the progress of the AI product of each company. So, how's that product gonna um make the progress of the AI in terms of a company? Then we look at like the readiness. We call it like, is the business model gonna be ready for the disruption? So those two scores
13:00
Unidentified speaker (speaker_3): is really from the partners we work with, like AI native researchers. They give like that score to select individual companies. Also, we have allocation cap for each layer of the allocation, the hardware infrastructure and applications. there's a 40% cap. And that that weight will change along
13:20
Unidentified speaker (speaker_3): the time as where the value is going to be created. With those two things together, we're going to have achieved this might value migration within the AGX portfolio.
13:29
Unidentified speaker (UNKNOWN): All
13:29
Unidentified speaker (speaker_2): right. So like I said earlier, Elon is saying, or he did factually say that he thinks that I don't know if you give gave a time frame, there could be more humanoid robots than humans. Where are we in the humanoid story?
13:42
Unidentified speaker (speaker_3): Yeah, I think like it's more than humanoid. Um, we have we have humanoid company called Aptronic, but like we which we feel like the physical AI model can do so many things. They can control vehicles, they can control trucks, they can control industrial arms. So those are the companies that
14:03
Unidentified speaker (speaker_3): they previously is solving engineering problems, is solving optimization, but now the same framework they have been using to train large language models over the last 10 years can be used to train the robotic model, the physical AI model, to let AI understand the data from the real world.
14:24
Unidentified speaker (speaker_3): That progress is like happening right now. So we have investing like general intuition. They have data from, say, the games, they have so many data like access. So they they can real world interaction can be used for data training, the bottle training, to control the robots. So we see
14:44
Unidentified speaker (speaker_3): that something, if that's kindly kind of like the current bottleneck for the humanoid, for the physical AI. The locomotion, the the hardware is already so van, so advanced. If you look at like the the marathon or like Kung Fu, you don't worry about that. You worry about intelligence. But intelligence is now breaking through,
15:04
Unidentified speaker (speaker_3): is at the turning point. Uh I think Elon Musk he he always had this perfect idea. It's like I want my product to be perfect. So that's why he always delayed on delivery. But when you have intelligence improving, not not just from Tesla, but so many other players look at like physical intelligence, skid AI,
15:23
Unidentified speaker (speaker_3): general intuition, the fifth release world app just got bought by AMD. The intelligence improvement gonna make a big difference here.
15:31
Unidentified speaker (UNKNOWN): So
15:31
Unidentified speaker (speaker_0): Derek, do you think this is f this is mostly in the manufacturing world where this is gonna going to see the biggest change? Are we gonna going to see this robotics elsewhere? Like, where is this first going to hit when it happens?
15:42
Unidentified speaker (speaker_3): Yeah, I think first is in manufacturing industrial. Um, that like Apptronic, they're based in Austin. They work with like DeBelle, they work with so many like manufacturing older companies locally. I would say that something is um This low low-hang hang fruits because that's tangible, high value, um, and it's
16:02
Unidentified speaker (speaker_3): environment. that's really safer in a way for robotics to navigate compared to our home, compared to like cell receptor, you have to deal with kids, docs, and I think like you need a higher intelligence level and safety level. You need like GPT 3, 4, or 5. Think about that. You need a
16:21
Unidentified speaker (speaker_3): further safety to deploy that into the home environment.
16:26
Unidentified speaker (speaker_0): I told Michael, I'm not buying the first robot because if the robot kills you, that's on you. You got to wait for later ones to come. But isn't this just going to require that much more capital as well? Because the physical world is so much more intensive. Like the fact that AI can solve like mathematical equations and software makes a lot of sense to me. It's like it's like filling in a puzzle. But having
16:46
Unidentified speaker (speaker_0): robots in the real world, that's a whole nother set of money that's going to be raised. gonna Is that like what's a feasible time horizon here for this stuff?
16:55
Unidentified speaker (speaker_3): One thing is look at like OpenAI's latest model, the Astra, many like robotic firms are using that as part of the brain. So the trillions dollars spent on the data centers and training the kind of come like digital AI models, those intelligence can be useful for robotics. So we're
17:15
Unidentified speaker (speaker_3): not gonna reinvent in the wheel. We're not gonna invest in another trillion dollar just on train like data centers and the models. You can use all that. But you need additional capital for a different things. You you need different capital for the the moving part. You need a body. You need actuators to move things
17:35
Unidentified speaker (speaker_3): around when you have the intelligence to control. So I think that's more what are different opportunities set uh in the industrial world to have those components ready, have those um machines ready requires different set of capital and different investors.
17:51
Unidentified speaker (speaker_0): My biggest question for you is like, how do you manage the the differences between investor expectations changing and in your portfolio? Because we're seeing is something something like semiconductors now, where the earnings fundamentals have been off the charts good this year. But like these companies are in a pretty swift drawdown right now, right? Like the semiconductor space is
18:11
Unidentified speaker (speaker_0): 40% off the highs or something. It's still a great return for the year. But how how do you manage that where the expectation game and the fundamentals are constantly in conflict with one another, seemingly? Where you can have these wonderful fundamentals, but expectations are so high that sometimes the stock stocks can underperform those expectations and those fundamentals. So how do you how do you deal
18:31
Unidentified speaker (speaker_0): with that potential mismatch?
18:33
Unidentified speaker (speaker_3): Previously, there's a lot of mismatch. We have like TMS TSMC just deliver like phenomenal revenue growths, further caps guidance, and they check the data center build out. They have very strong confidence on the delivery of those data centers. Even with that,
18:49
Unidentified speaker (UNKNOWN): the
18:50
Unidentified speaker (speaker_3): equity was down. post-earnings. Because of the rumors that Enthropy is gonna, their revenue can potentially slow down if you remember this like a few weeks ago. That's that's kind of like there's all kinds of rumors can create like a lot of uncertainty because people thought, oh, like I'm so concentrated trading in this,
19:09
Unidentified speaker (speaker_3): uh, what if the AI market can slow down? You can't can like have that. And I think that's inevitable.
19:15
Unidentified speaker (speaker_3): when you have seen something like AI, because the change is so dramatic and the capital behind it is so large. So the volatility tends to be higher right now at this stage, even though the fundamentals are really strong in supporting that. So for investors, how do you navigate that? Like you have
19:35
Unidentified speaker (speaker_3): to really size it right, right? Look at your risk appetite uh, with your risk budget. Then like in terms of AGX, we look at like things we have really, really like um rule-based methodology on our public equity portfolio. This 40% like cap on the hardware infrastructure applications,
19:55
Unidentified speaker (speaker_3): really like limiting our exposure to overconcentrated overconcentrate it uncertain in certain opportunities compared to many competitors. I've seen some ETF they're talking about AI, but like they're like 70% in hardware. Yeah, they're like sanitrate. So it's something potentially giving more like volatility to investors compared to a more diversified
20:15
Unidentified speaker (speaker_3): approach. So I think like, yeah, investors have to really do the research themselves to understand what's what's inside the portfolio.
20:21
Unidentified speaker (speaker_2): Is there a hard cap? So you've got 7% of the top. seven private holdings as a percentage of the fund. Is there a hard cap in terms of how much private assets can be in an ETF like this?
20:34
Unidentified speaker (UNKNOWN): So
20:34
Unidentified speaker (speaker_3): ACC guidance is um lower than 15%. But as we invest, we target about 10% to have a 5% buffer because there could be appreciation and there's some other things. So we talk about like uh 10% illiquid assets
20:50
Unidentified speaker (UNKNOWN): to
20:50
Unidentified speaker (speaker_3): manage liquid assets. liquidity. That's really like something for like institutional investors, they tend to allocate that much to the private as the like a sleeve of portfolio. Um, so we think for just everyday investor, that's appropriate as well.
21:04
Unidentified speaker (speaker_2): You mentioned that you you're on the uh the enthropic cap table. Is that the case with every private investment that you hold, or are there other ways for you to obtain to these shares? or the exposure if it's not direct investments?
21:18
Unidentified speaker (speaker_3): Yeah, so there's usually like two ways. One is through like direct investment, right? Either through primary financing, direct on cap table or tender over round, some secondaries from existing shareholders. So far we're just direct investment. Then there's SPV as well. You can use like a video
21:38
Unidentified speaker (speaker_3): codes code through like a counterparty to get some like second-tier access. So that's like I think there's some counterparty risk and some companies really not happy with that. So we haven't really used um SPV in our current investment. So that's kind of like the difference. For investors, I don't think a lot of like investors
21:58
Unidentified speaker (speaker_3): they understand the difference. But uh we talk to our a lot of institutional investors. They feel getting to a cap table is the most transparent way and the easiest way to express this um allocation.
22:09
Unidentified speaker (speaker_0): I know you guys have a lot of rules-based systems, which probably helps, but how much turnover do you think this fund is going to have in the years ahead? Like, do you think that there are still other brand new companies out there that aren't even on the radar yet? Or do you think that these companies that are at the forefront are pretty much entrenched at this point? Like, how much turnover in the leadership do you think is gonna be going happening to in the years ahead?
22:28
Unidentified speaker (speaker_3): I think like the leadership is getting more concentrated. Just think of the model space, right? It's getting more concentrated in just two players from like 10 10 plus players. But also the dynamic is also changing. Think about like Anthroopic this year, they have some moment where they're losing some market share to open source
22:47
Unidentified speaker (speaker_3): model. But if you look at like recent data, they're gaining market share. back. So I think the dynamic is gonna be more hybrid. That you have several leaders, but the challengers can probably gain some market share and secure like a some value in in other like low-cost jobs or like lower
23:07
Unidentified speaker (speaker_3): value jobs. So for the model side, that's gonna be the case. For many other categories, I would say like in terms like physical AI, there's no leaders yet. There's so many emerging players. It's almost like AI in two 2022, 2023. So so much potential. And we're betting on early
23:27
Unidentified speaker (speaker_3): players that we feel comfortable. That's a totally different like market that that the the the leadership can change rapidly.
23:36
Unidentified speaker (speaker_2): This might be outside the realm of like what you typically talk about, but the SP 500, for example, is obviously heavily skewed towards this theme already. So how are investors coming to you and asking, like, hey, where does this go? Do I take this from the SP 500 bucket?
23:53
Unidentified speaker (speaker_2): If I add this, am I like all in, all in?
23:56
Unidentified speaker (speaker_3): I think like a lot of investors they've seen like SP as where it concentrating in a few names, but when you look at like a long-term allocation to AI, it's not likely to just like concentrating a few names because the value can migrate to other names as well. Using a dedicated allocation as a satellite
24:16
Unidentified speaker (speaker_3): position to a broad AI ecosystem, also like with the private allocation, I think that's something where very it's strategic for them as well, as a complement
24:24
Unidentified speaker (UNKNOWN): to
24:25
Unidentified speaker (speaker_3): the SP S P rather than repeating. replace the S P. But you have other like exposure coming out of S P. The S P leaders kind of like are the winners over last decade. But the the winners for next decade can be very different. So have a dedicated allocation to AI to have a more balanced reel on
24:44
Unidentified speaker (speaker_3): those names Games is kind of like a long lone-turn term approach or for many manning investors. lasters.
24:48
Unidentified speaker (UNKNOWN): I
24:48
Unidentified speaker (speaker_0): I'm am curious where you see the whole autonomous driving. fitting into all of this. Because that to me is the easy logical step for something that's relatively robotic and AI can help. It first took longer than we thought to get here. You know, 10 years ago it seemed like autonomous driving was going to be like the next big thing immediately. Now it finally seems to be here. How much market share can autonomous driving gain?
25:08
Unidentified speaker (speaker_0): Or is it just going to be sort of the robo robotaxis taxis at first?
25:11
Unidentified speaker (UNKNOWN): I
25:11
Unidentified speaker (speaker_3): think robotaxi is definitely the first adoption. Think about autonomous driving, we have been talking about that for many, many years. But I think the challenge for that is the economic, the uniconomy really doesn't work. But now we're in a point that deploying autonomous driving as a
25:31
Unidentified speaker (speaker_3): robotaxi um is gonna make it the RI is very high. So if you have further deployment, that is really like economic.
25:42
Unidentified speaker (UNKNOWN): So
25:43
Unidentified speaker (speaker_3): that's the big difference here.
25:45
Unidentified speaker (UNKNOWN): So
25:45
Unidentified speaker (speaker_3): if you look at Waymo, if you look at like Neuro, if you look at like everything like Elon Musk, there's so many players right now, they see this opportunity is happening. So that's why over the last year,
25:58
Unidentified speaker (UNKNOWN): I
25:58
Unidentified speaker (speaker_3): think it's kind of like a turning point for the the unit economy for the society. this autonomous driving. So the going forward in the next three to five years,
26:07
Unidentified speaker (UNKNOWN): the
26:07
Unidentified speaker (speaker_3): penetration rate is gonna be very different compared to the last 10 years.
26:11
Unidentified speaker (UNKNOWN): Derek,
26:12
Unidentified speaker (speaker_0): for people who want to learn more, where do we send them to learn more about um AGIX and all of your different funds?
26:17
Unidentified speaker (speaker_3): Yeah, so for AGX, people can check craneships.com slash AGX. We have published all our holdings, uh including our private holdings. We have white papers and research piece uh on the AI and all on the private holdings as well. Feel free to check AGX looking at our website.
26:35
Unidentified speaker (UNKNOWN): Perfect.
26:35
Unidentified speaker (speaker_0): Thanks, Derek. Appreciate the time.
26:37
Unidentified speaker (speaker_3): Thank
26:37
Unidentified speaker (UNKNOWN): you.
26:40
Unidentified speaker (UNKNOWN): Okay,
26:40
Unidentified speaker (speaker_0): thanks to Derek, and thank you to Crane Shares. Remember, check out CraneShares.com slash AGIX. Learn more about the public-private AI and technology ETF and email us Animal Spirits at the compoundnews.com.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.

More from Animal Spirits Podcast

Animal Spirits Podcast

What the Hell Just Happened?! (EP. 490)

Episode discussion: Speakers analyze market volatility, AI infrastructure backlash, and political impacts on tech valuations.

Speakers’ claims: The hosts discuss recent market volatility, attributing sharp drops to hedge fund degrossing and leverage rather than fundamental shifts. They examine the political backlash against data centers, noting voter rejections of ballot measures and the potential for increased build costs. The conversation covers geopolitical risks to NVIDIA following personal scandal allegations and Amazon's aggressive AWS growth targets. Finally, they compare current market conditions to historical periods like the dot-com bubble, highlighting fixed income opportunities and labor market anomalies.

Market VolatilityAI InfrastructurePolitical Risk
1h 13m5 assets25 claims to verify
Animal Spirits Podcast

Talk Your Book: Customized and Income

Episode discussion: Dimensional and NEOS discuss customized SMAs and option-based income strategies for tax efficiency and yield

Speakers’ claims: Kacie Walsh Jason of Dimensional Fund Advisors explains how separately managed accounts (SMAs) offer superior tax loss harvesting at the individual security level compared to aggregate fund harvesting, claiming up to 10 years of tax alpha. Wes Mathews of NEOS describes the growth of option-based income ETFs, which sell upside participation to generate yield, noting NEOS has grown to $36 billion in assets. Mathews details specific products like QQQI for high yield, XQQI for leveraged income, and QQQI for downside protection, while emphasizing the convenience and risk management provided by professional managers. The speakers highlight that these strategies cater to older investors seeking income and tax efficiency in mature portfolios.

Tax EfficiencySeparately Managed AccountsOption Income
33 min3 assets20 claims to verify
All insights