Is Robotics The Next Megabubble?

1. From Crypto to the Robotics Mega Bubble

Andrew, thank you for joining the 1000x podcast today. We are super psyched to have you. You're working on some really interesting things. Welcome.

Thanks for having me, Avi.

This is fun. I know we've known each other for a bit, but we've known each other from the world of crypto, and now you're known as the Robo King—the moniker that people are giving you online.

I haven't heard of that one, but that's fun.

I've seen that around. You've gone super deep into the world of robotics, and this is a world that I don't think a lot of our listeners, or a lot of people, have really paid close attention to. It's getting hot right now.

I wanted to start by asking you: You started off in crypto. You started off as an investor, and then you got into robotics. How did that take place? Maybe walk me through your entry into investing and your evolution from crypto investor into robotics investor.

I would say my professional entry into investing was probably around the 2018 time period. I went really heavy into crypto when the markets were dying, and I would say I had a little bit of a Solana run, being early in sectors like DeFi that weren't really established at the time but eventually took off.

I brought on 5 or 6 other people to help manage the portfolio and start what is known as Mechanism Capital. We never took outside capital, and our primary focus was investing and trading the crypto markets, as well as active venture capital investing. We never really ventured too much outside of the crypto space until 2023, because that's where we felt our expertise and area of competency was.

In late 2023 and early 2024, a friend told me about the company Figure AI. I watched a video of the founder, Brett, about why he believed humanoids were going to be the future and what they were building with the company, and it just made sense.

Humanoids were always this sci-fi dream, right? You'd see them in every sci-fi movie, but you didn't have them in real life. I think the concept has been around for at least 100 years, where people have imagined machines that are shaped like humans and can traverse and manipulate the world like we can.

There were too many technological barriers. The hardware design is difficult, yes, but it was solvable. To make a robot understand the world in the same way that we do and be able to interact with it in an infinite amount of different circumstances and environments, with different objects, requires a very high level of physical AI.

With ChatGPT coming out in 2022, I think it was clear to the world that we were going to have essentially digital AGI within a reasonable amount of time. You could extrapolate that to say we're going to have physical AGI as well.

From that perspective, I think it was clear that there was an inflection point that was going to occur really soon. Robotics development, or humanoid development, was looking like this, and now it's going to be like this in terms of what these machines are actually capable of.

The best time to invest, in my opinion, is at these inflection points, because not everyone realizes that the pace of development is going to change, the expectations for development are going to change, and the market is going to reach that future state a lot sooner than everyone would think. The market's not always pricing it that way.

It reminded me of crypto in 2014 and 2015. Some early adopters were starting to understand that it was going to be really big. There were a lot more resources going in to build the industry, and it was really the perfect time to jump in—not just because things were underpriced, but because the field was complex and not well understood.

That's exciting to me as an investor: being able to tackle something where there's no framework for evaluating it, like there is for SaaS companies. That's really established. You have to figure out the best way to evaluate it yourself, and if you can do that, you have a tremendous edge over all the other investors in the industry.

I felt like we did that in crypto and DeFi back in 2019. This was very similar in that sense. It was so interdisciplinary, and there were so many different complex topics involved in robotics. It's going to be a challenge to learn about all of them and evaluate how important each of the different factors are in robotics development and in building a company in the space.

2. A $10 Trillion Industry Hiding in Plain Sight

We felt like we could really differentiate ourselves and establish a foothold as one of the top robotics investors in the world. The timing is important.

That makes total sense, but I want to dig in here because I think this is something that's really important, especially for investors in general and people who are looking at how to even begin to navigate this world.

You're in crypto and DeFi. There are so many different industries that you could have gone into. You could have looked at space, AI, or robotics. There are probably some red herrings out there—industries you could have gone down that didn't pan out.

What really drew you specifically? A lot of the things that you said could be said about space as well. Were you watching I, Robot? Was there an inspirational moment where you thought, "Robots"? Have you always been interested in this? How did you actually end up here? Maybe talk through that process.

I wouldn't say I was a super sci-fi nerd, but I'd always appreciated some of the movies. Space wasn't as hot as it is right now, but SpaceX was already starting to prove itself, I think. It was at a significant valuation, and there was interest in the space.

If you look at the total market cap of all private robotics companies at the time, it probably was around $20 billion. Right now, it's probably around $100 billion to $150 billion, maybe pushing $200 billion.

XRP is close to $100 billion. The total market cap of Pokémon trading cards is around $50 billion. I think that puts into perspective how early we are in this industry, which I think is going to be worth tens of trillions of dollars. The industry as a whole still has potentially 100x, or even 1,000x, upside remaining.

It seems a lot more greenfield and tangible as well. Space is cool to look at, but I don't think most people are going to ride a rocket ship. Maybe they will in 20 years, but robots are going to permeate everyday life, and that's going to happen a lot sooner.

What do you think about that? Let's say the world's in 2035, or whatever year you think the robots are coming. What types of things are these humanoid robots doing? How are they contributing to this world? How might you and I interact with them?

3. The Companies Building Physical AI

Man, they're going to do everything that humans do in the real world. They could be your personal assistant. Right now, it's pretty expensive to have a personal assistant, but if you can have a humanoid robot that costs, say, $20,000 for an economy one or $50,000 for a premium one, that is pretty cost-efficient, especially if you're doing that as a one-time cost and maybe paying a little bit on an annual basis.

That can add a tremendous amount of value to people's lives. In America, a good personal assistant could cost $40,000 or $50,000 a year on the lower end of the range. I've seen executive assistants go up to $200,000 or $400,000 in all-in compensation.

But not just that—you can have robots taking care of your parents.

That'd be amazing to have. You can have robots working in factories. You can have robots driving cars. Even though we'll have autonomous cars, we're still going to have manual cars as well. You can have robots working at hotels and restaurants. You can have them go on the moon as well, go to space, and be in environments where it's difficult to have humans operate.

The first robot astronaut. We're really hitting the sci-fi now. This is crazy. I'm curious as to what you're seeing right now, like the general state of the robotics market. Talk to me about the top companies and what they're trying to accomplish. What are they building?

I know you have investments in Apptronik and Figure. What is being built right now? Is it everything that you're describing, or are we starting a little bit lower?

There are companies that are building toward basically solving general-purpose robotics. Those are companies like Apptronik and Figure, because they're focused on the humanoid form factor, which is the most versatile. That means they can go really anywhere a human can go, and even more—to space.

Then there are really exciting companies that I feel are less hyped right now but are going to be real giants in the future of industry. We've always had industrial arms. We've had automation, we've had machines, but the ability for them to be intelligent and not need weeks of programming and tens of thousands, maybe hundreds of thousands, of dollars of engineering cost just to set things up opens the space up for you to deploy robots in so many more applications around the world.

Not everything has to be moving like a humanoid. A lot of stuff done in factories is stationary, and so you can just have an industrial robot arm. Standard Bots, one of the key companies in our portfolio, builds mostly cobots, which are basically smaller versions of standard industrial arms that can work safely among humans. They build industrial arms, and they also build some other types of general-purpose robots, but they're general-purpose in the same way that humanoids are.

That is a tremendous market. If I think there are going to be billions of humanoids, then there are probably also going to be billions of mechanical arms around the world, whether they're making coffee, cooking, or unpacking things in a warehouse. That market is huge.

If we think we're going to reindustrialize America and start manufacturing a lot of stuff here that was formerly manufactured in China, it makes sense to have a U.S.-based industrial arm manufacturer. There's only one that really exists at scale, and that's Standard Bots. There are a lot of companies, by the way, in Europe and Asia, but it's pretty key to have one that is homegrown, vertically integrated, and not reliant on China as a key component of its supply chain.

4. Why VCs Are Pivoting From Software to Robots

That's just one example of some of the other less-hyped companies in the portfolio that could be getting a lot more attention in the future.

This is an important point here. The market seems to have expanded not just in terms of market cap, but in terms of the companies that are entering the space. If you go back to when you first started, how has the market shifted over the last 2 years?

Are you seeing a lot more interest? Are you seeing traditional VCs come in now and start to get really excited about robotics, or is this still a niche area? Are these companies raising easily now, or is it still hard for them?

I would say that interest has really started to inflect over the last 3 or 4 months or so. The reason why it started to inflect was because there was this huge re-rating in software. There was this huge wake-up call earlier this year that some software businesses were not as durable as we thought they were. The perception of the prospects for future software businesses that are venture-fundable changed significantly.

In the VC industry, software is a majority of the investment interest. A lot of VCs had this existential moment where they were saying, "Hey, look, if we can't invest as much into software, where are we also going to invest?" Obviously, they were already looking into AI, but robotics and physical AI started to become more of an apparent next stage for the venture-capital industry.

People started to understand how big this market could be. There were a lot of concerns previously around CapEx costs, competition with China, the ability for us to manufacture in America, and just overall, VCs like to be a little bit late hopping on trends. That trend wasn't established yet, and so that changed in the last few months.

We're seeing some uptick in interest, but I don't think it's anywhere near what it could be. The amount of interest we have in AI companies right now—that's the equivalent, I think, of what it will be in the next 1 or 2 years.

You're seeing that start to trickle into basically every top-tier, Tier 1 VC and, I would say, Tier 2 VCs looking at robotics right now. Some of that is turning into term sheets at valuations multiple times higher than these companies were raising at just a few months ago. I think that's really going to accelerate.

5. The GPT-3 Moment for Robotics

That's exciting. Do you think part of it is going to have to do with the actual capabilities of robots improving? From my perspective, some of my interest came from Figure starting to livestream the robots working. When I saw humanoid robots actually doing things, and when you see videos of Chinese humanoid robots dancing, it just seems like the world is catching up to the fact that these things are actually happening.

I'm curious what you've seen, because I'm sure you've visited the factories and gone behind the scenes. Is there stuff that you've seen that's blown your mind yet? Are we close? Are these humanoid robots going to—are we going to see sports leagues with them? What have you seen that you find crazy with these robots?

I think we're pretty much at the GPT-3 level of robotics in terms of intelligence right now. I just think that it's not as striking or as obvious to people, because for GPT-3-level intelligence in LLMs, it was wrong half the time, and that's not impressive for a robot. For it to pick up a cup accurately only half the time is not impressive. But it means we're getting really close to solving general-purpose robotics.

What I would say is that some of the frontier models are really impressive because they're generalizing a lot more. That is kind of the hallmark of what you want to see in AI models that actually work: that they can face environments that they've never seen before. You put one into a completely new home or factory that it's never been trained on, and it can operate in it in the same way as if it were in the factory or home that it had training data on.

It's also able to interact with new environments. It's able to pick up a new package that it had never seen before. One of the key capabilities of the AI models of the last few years was this concept of in-context learning: the fact that you can say something to an LLM, and it would understand that context and then be able to give you an answer. You would be able to have a conversation with it, as opposed to it forgetting every single time.

We're starting to get that in the physical AI models as well. I can show it an example of a simple task, like, "Put this item into a box." It's able to do that. That's a really simple task. It's called pick-and-place.

We can do that, but now it's starting to expand to more complex, long-horizon tasks as well. Maybe tasks that might occur in factories, like, "Hey, look, put these items into a box, package it, and then take that box and put it over there." The more complex these tasks get, the more it can start to replicate what humans are doing in the real world.

Not only that, the more recent models are starting to show that they have the ability to memorize things internally or innately. There was an example shown by Rohde AI where they played the shell game. They had 3 cups, put a ball under 1 of the cups, shuffled them around, and asked the robot to pick which cup the ball was under. It was able to pick that up.

A lot of these capabilities or characteristics can be called emergent. As you scale up data and scale up compute, these models are capable of more and more things that allow the robot to function more like a human would be able to.

Unlike a traditional robot, which is very preprogrammed to do 1 thing, if any small thing changes—like I move this cup from here to here—it just doesn't work anymore.

We're getting really close, I think, to the ChatGPT-5 of robotics. If we think about the time from GPT-3 to GPT-5, that was 2 years. I think it's going to be way less than 2 years—something like 1 year.

The reason is that there were a lot of learnings from a research perspective going from GPT-3 to GPT-5. It wasn't just scaling compute. It was learning how to best do RLHF, how to best structure mid-training, how to find the best mix for pretraining data, and how to create infrastructure to annotate data more effectively and efficiently.

A lot of these learnings have already been made—we've figured out a lot of these things. Now we don't have to replicate all of them again; we can apply them to physical AI models to compress that timeline. Robotics—and AGI, I think—is going to happen a lot sooner than people think.

So, I'm curious about the edge that these robotics companies have because you've talked a lot about the intelligence models. Is it possible that your Claudes and your ChatGPTs of the world, if we achieve AGI on that front, can just be hooked up to basically any old hardware?

In that case, where does the edge remain for the robotics companies? Or are they building very specific models, and do you think it's going to be very difficult to translate them?

You can't just hook Claude up to a robot and have it work. The difference between LLMs or VLMs and robot foundation models is that robots need to understand the physical world around them. Traditional models like Claude Opus understand the world based on text.

Imagine you had a blind person who could only hear things, and suddenly you gave them vision. They had never moved around—they were a blind person with no limbs.

I mean, that sounds like a really sad person.

Now imagine you gave them limbs, gave them sight, and told them to interact with the real world. It wouldn't work. Maybe they could figure it out after a lot of trial and error, but it's not going to work out of the box.

I make that analogy to help you understand the difference between these LLMs, which only understand the world through text, and robot foundation models, which need to understand feeling. I need to be able to understand, “Hey, look, if I look at this, I know how much pressure I need to apply to keep it held and not drop it.” Purely looking at it, I guess some vision-language models can understand that it's being held, but they can't understand how much pressure I'm applying.

That's a human experience that also needs to be translated into robot models. There's this kind of further semantic understanding that robots need to have. If I just say, “Hey, take this cup and put it on the drawer over there,” it's not very obvious for an LLM which drawer I'm talking about or what I mean by “put it on top.”

If I'm carrying a glass of water, I need to understand that I don't carry it like this because the water can fall out. I need to understand physics as well. I need to understand that if I interact with an object too hard or hit something, it might damage me or the object.

That kind of understanding isn't imbued in something like Claude. It needs to be imbued into what robot models are—what people are working toward for robot models today. For example, if I'm picking up a bottle, how do I know if I should pick it up with 1 hand or 2 hands? If the bottle gets big enough, I need 2 hands, but at what size does that make sense?

There's a lot of that intuition. In robotics research, they call that priors or affordances—things we take for granted as humans. Some of that we're born with, and some of that we learn as we grow up. Robot models also need to learn it.

6. How To Get Smart on a New Industry

They need to be able to learn by themselves as well, because there could be new situations we put them into that we haven't even thought about. They need to be able to adapt to them like humans can. That's kind of the next level of robot learning.

It really seems like you've done a ton of work to understand this field. I think a lot of our listeners right now are people who are thinking, “How do I allocate my money? How do I invest? How do I spend my time?” Most of our listeners are investors or traders, and they're in the markets.

I think you've done maybe one of the most impressive jobs of coming up to speed on a new industry that I've seen in a very long time. I'm curious about your approach to understanding robotics. Once you decided, “Hey, I'm going to go invest in this thing,” how did you even get up to speed?

Were you getting on the phone? Were you calling people and talking with them? Were you reading books? You can't really read a book on this kind of stuff. Maybe you can, but what was your process for becoming what you are now—one of the world's experts on this industry?

You've invested in a ton of different companies, and you're very close to the ground. How did you get to the place where you can just talk to me for 27 minutes so far about robotics?

It was pure obsession. For the past 2 years, it was the only thing that I thought about from day to night, on weekends, and so on. There were just no days off. The only time I took off was to surf a few days a week and work out a little bit.

Well, I assume you'll be able to socialize with the robots at some point. That's part of it.

Yeah, I just talk to Claude. That's my primary point of contact with another being, which is really helpful, by the way. To answer your question, in terms of understanding some of these concepts, there are a lot of experts we've become friends with—founders and really great researchers—but I don't want to bother them all the time.

If you go through a paper, it can seem really daunting because there are so many technical concepts. Now it's so much easier because I don't need to read a textbook. I just tell it what I don't understand: “Explain this to me. Explain this concept to me. How is this ablation experiment done? Why is this model architecture architected the way it is? What are the trade-offs with other model architectures?”

I think a key part of investing is honing in on and focusing on what you don't know, identifying those unknown unknowns, and filling that gap. It's easy to say, “I don't know something. Forget it,” or to focus on what you do know. But I think it's really important to become the best investor that you can be, to have vision everywhere, and to try to figure that out as best as you can.

I think I've done a decent job over the last 2 years, but nothing replaces decades of experience as well. That's why we talk to other experts in the field and why we built a team at RoboStrategy that has robotics experience across a lot of different domains.

For example, Scott Walter on our team has had 40 years of robotics experience. That's more experience than some teams we talk to have. He's built 2 companies and sold both of them in the space, the last one to Kuka, which is a very large robotics company.

He's had clients across aerospace, consumer goods, and a lot of the areas in the physical world where robotics are going to be applied. He understands the pitfalls and the things we should be paying the most attention to. That's a key part of investing as well.

There are thousands of variables that you could look at. Which are important and which are not, and how do you weigh the importance of each? You have to create this mental model over time, which is not simple, and there's no textbook formula for it. You do it through experience and debate, and by going down really deep rabbit holes of thinking: If this happens, what are the outcomes for the company?

You do that across a thousand different mental exercises. If I were to zoom out a little bit more, investing, from a first-principles approach, comes back to how I analyze the risk-reward of something. That sounds really simple, but 2 different people can come up with widely different estimates of the risk and reward for a specific investment or trade. Someone can think there's a 99% chance of something being successful, and someone can think there's a 1% chance.

Mhm, and someone can think, “Hey, look, the risk is I lose all of my money.” Someone can think it’s like: 50% of the time, I lose all my money. Someone can think, “Hey, look, I lose 20% of my money a tenth of the time, and then 50% of my money a quarter of the time.”

And so you have to think in probabilities, right? To think in probabilities, you have to understand all the different scenarios that can play out. To understand all the different scenarios, you need really deep industry experience and just thought that’s put into what can happen. And then, to do that, you also have to understand—it’s kind of like this meta thing—you have to understand your own understanding of the space, right?

Because if I overestimate my understanding of the space, then maybe I won’t dig into it enough, and I will have a wrong estimate of what the real risk-reward of this investment is.

How do you check yourself on that? I mean, how do you make sure you’re not overestimating? What’s your approach to that? Are you surrounding yourself with people who are arguing with you? What’s your process there?

Our culture as a firm is just very truth-seeking, very reality-seeking. We try not to be dogmatic. We can get very high conviction about certain investments, but only because we’ve thought it through so deeply, talked to so many experts, and done our due diligence.

How do you not overestimate yourself? I think it’s maybe having a little bit less of an ego. We all have egos. I definitely have an ego. I think just prioritizing truth-gathering helps with that, and also constantly trying to understand: What can I be missing? What perspective am I not seeing? Where can I be wrong?

If you’ve validated 5 different avenues of where you could be wrong and it turns out, “Hey, I’m not wrong. Maybe the market’s wrong,” then, for example, I think that’s what happened with Figure AI. Okay, then I can feel a little bit more comfortable about my understanding of the investment and where it’s going to go. But that is really important.

7. The RoboStrategy Playbook

I think maybe a lot of investors overestimate their knowledge of the field, so you just have to be really obsessive in turning over every single stone that you can.

That makes sense. I think your process obviously led you to a really great spot. You were very successful in crypto. Now you’re venturing into new ground in robotics. I want to talk a little bit about how you actually are supposed to invest.

So we basically talked for 30 minutes about what the opportunity is. The opportunity is clearly massive. And now the question that’s probably on everybody’s mind is: How do I actually invest in robotics? All these companies are private, and you seem to have come up with some solution to this.

You launched a public company called RoboStrategy, which now trades on the Nasdaq. It basically rolled up the investments that you had privately into a publicly owned vehicle so that people could buy it. And I want you to talk, if you could, a little bit about how you decided to do that, because most people, I think, would just want to basically hold. I mean, why are you letting other people buy your portfolio now if you think this thing is going to the trillions? How did this all come about?

I think it’s going to be really hard for people to make their own decisions about investing in the best robotics companies. It’s going to be near impossible, I think, to do it to the same level of excellence that we’re doing it at, because we have a team of robotics experts and this is all we do day in, day out.

That’s not just from an experience point of view, but also from the network that we’ve built in the space to be able to diligence everything from talent internally to supply chain. Are these companies really ordering as much as they say they’re going to order? Do they have those relationships? Because founders, a lot of them, are great at giving pitches, but maybe not all of the claims can be validated all of the time.

Maybe someone looks really great on paper—they have a PhD from so-and-so university—but in operating within a company, maybe they’re not the best person for that, and it won’t translate. There could be hundreds of hours that go into diligencing a large investment. And that doesn’t even include the cumulative experience that’s taken place.

We’re going to be able to talk to customers that people aren’t going to be able to talk to, to see: “Hey, look, are these robots really performing as well as the company claims? Are you really going to be buying a lot more robots in the future?” That’s why we built the firm, because I couldn’t do it alone, right? And no, nobody can do this alone. That will result in better decisions being made on investment choices and sizing of the investments.

I think the people who are trying to chase some of the hype—we hear about some of the investments that are being made, and it feels like people are trying to play catch-up a little bit, right? They’re trying to find, “Hey, look, I want the next Figure AI.” But they just haven’t seen, or maybe thought too deeply about, all the ways that it can also not work, because robotics is a really hard industry.

Even though there are going to be a lot of winners, there are going to be some really big winners. It’s still undecided how much breadth there’s going to be—how many winners there will be. There are going to be a lot of failures, right? Just in venture investing and startups in general, there’s going to be a very, very high failure rate.

And so, going back to your question, why did I decide to start RoboStrategy instead of continuing to invest privately? It’s because I started talking about these robotics investments on X, and I probably got over 1,000 DMs and replies from people asking me, “How do we get exposure to the space, and what companies should we be buying?” In the public markets, it’s really just Tesla and maybe 1 or 2 other companies.

Tesla is a $2 trillion market cap. I think it’s going to do great, but all of the exciting stuff—and I think there’s going to be trillions of dollars of value creation happening—is in the private markets. Most people won’t be able to get access to that space. SpaceX, Anthropic, and OpenAI are going to be at $1 trillion-plus before they go public.

Look, my primary goal is winning. It’s not like I’m an altruistic person. My primary goal is—I want to give everyone access. I think that’s a really nice factor in terms of what we’re building. And it’s important, right? For RoboStrategy to work, the unique value that we provide is that, in the public markets, if you want exposure to this growth that’s going to happen in the robotics industry—from maybe $100 billion to tens of trillions of dollars—there’s no other way to get it.

For us, we always focused on early-stage investing at Mechanism Capital because that’s where we felt there was the most asymmetry, and if we took additional capital, it would dilute our returns. You saw with OpenAI and Anthropic that now the power law can really apply, and we believe it’s going to apply for robotics as well.

I think economies of scale are a really significant factor, because if you don’t have scale, it’s going to cost you what? Like $500,000 to $1 million to produce a robot just from a BOM perspective, right? Not just from the R&D perspective. Whereas if you have scale, you can produce it 90% cheaper.

With the understanding that scale is important and power laws are going to apply, we’re going to be able to put tens of billions of dollars to work in the Anthropic of robotics because there’s going to be a multitrillion-dollar winner. And it’s going to be a very high risk-adjusted return. Previously, you look at growth-stage investments, and you’re targeting maybe 20%–25%, right?

Mhm.

And now I think it’s reasonable to believe that even if something gets to a $1 billion valuation, I can target 100% year over year with what I would consider maybe a very similar amount of risk. It might not be the same underwritten risk from a different investor that doesn’t understand the industry as well. But from somebody who understands how this might play out in all the competitive factors and whatever, I can get a very high amount of comfort with some of these later-stage companies.

Maybe they’re not selling billions of dollars of robots yet, but I have a strong belief that they will in the future, and they’re going to be able to scale that to very, very high numbers. And so, how do I get there? We could do it in the private markets, but we tried really hard—maybe we can get to a few billion dollars within a year. But that’s not really interesting. If we want to do this, and we’ve built the best investment team in robotics so far, I want to do this at tens of billions of dollars of scale.

I think the market opportunity for an investment fund like this is actually in the hundreds of billions, at least for now—maybe more in the future. Elon says, right, that the public capital markets are 100 times bigger than the private capital markets. Traditionally, VC has never been done like this before. There are a variety of reasons why, but I think the time is right now to essentially invent an entirely new type of venture capital model that benefits both public-market investors and fund managers, and can scale in a new way.

Yeah, I want to also—there are 2 things that I want to make sure we hit on. Number 1 is, you named this thing RoboStrategy, and it seems to be an allusion to MicroStrategy in a way. There are a lot of these SPVs out there that have been trying to go public, and you just buy a piece and they have high fees. But you're doing something different, right, with the structure. Specifically, what's different about your structure that you think is going to be accretive? You're running, in some ways, the MicroStrategy playbook, taking the good and leaving out the bad, right?

I think MicroStrategy was definitely, I would say, an inspiration for our model, but they were actually not too different from some of the other similar models that had existed. They basically took the private-equity-to-public-company roll-up model and then applied it to Bitcoin. How that works, by the way, is you have a lot of these big companies in every sector. One example is TransDigm; another example is Constellation Software. For TransDigm, they focus on aerospace supplies; for Constellation Software, it's software.

In public markets, they trade at 15–40x multiples. For smaller private companies that are in the same sector, they might trade at 3–8x multiples, right? Different markets price the same asset differently because you have a different set of market participants. What they do is, their whole strategy for growth is predicated on M&A.

I buy this private company, right? Maybe I spend $100 million. Once it's on my books, it re-rates immediately to $400 million, and so it is immediately accretive. If I have a reasonable enough cost of capital, I can just do this on repeat over and over and over again. Instead of relying on organic growth within my existing companies, because I feel like that is maybe a little bit of a harder lever to pull for growth.

That's kind of what MicroStrategy did, right? I guess the difference with MicroStrategy was that there was a big access issue. There was huge demand for the stock because people wanted exposure to Bitcoin and they couldn't get it in the public capital markets, right? Even though you can buy it on Coinbase, a lot of funds, institutions, and investors have restrictions or can't access it. So they buy MicroStrategy for exposure.

Even after the ETF came out, I thought the premium would collapse, right? But it still stayed really strong for a while, and they were able to issue billions of dollars more equity because, man, it turns out that capital base that has these restrictions is really, really large, right? These RIAs, endowments, pension funds, et cetera.

If we think about access as really the primary demand driver, that issue is multiplied 1,000x for private robotics investments, private venture investments, and in general, right? It's really only this select group of venture capitalists and sometimes their networks that invest in these SPVs that can get into these deals. I think for robotics it could be even more constrained if we're really successful because maybe we'll be the primary financier of a lot of these rounds.

And not only that, but Bitcoin—we think about it as the TAM for MicroStrategy, right? The TAM for robotics is going to be multiple orders of magnitude larger. Bitcoin's market cap right now is below $1.5 trillion. Amazon is larger than that; Nvidia, Apple, and Google are all larger than that. AI is going to be as well.

I think it's very fair to say robotics can be as well. Not only do we have a better selling point because there's really no other avenue for the public markets to get exposure, but we have a larger market as well. I think we could be larger than MicroStrategy, right? Their peak market cap was around $100 billion. I'm reaching for something that's more like SoftBank scale—in the hundreds of billions.

You heard it here first: Andrew Kang is going to build SoftBank Part 2, but bigger.

A closed-end fund, right?

Yeah, I mean, it's true. It's true. And so, in order to do that, you're going to need to expand past robotics at some point. Or do you not think so? Why not? I mean, you think that you can capture, basically—I mean, let's say that the robotics market goes to $10 trillion. How much of that market do you think you'll be able to capture?

Well, $10 trillion is, what, 7x the current Bitcoin market cap.

Yeah.

And so MicroStrategy's market cap times 7, right? That's already in the hundreds of billions.

It's true.

And that's without the increased access, right? There are so many avenues for people to buy Bitcoin. And then you asked earlier what concepts we're taking in. We're taking in the accretive issuance as a way to scale and to have a second engine to generate shareholder value, right, to increase our NAV per share.

What we're leaving out is the leverage part of it. We're not going to do preferred shares. We think there's a huge amount of upside in our underlying investments because of how early it is in the market and how much we think the industry will continue to expand. If we use leverage, it will be very, very selective and we'll probably not have it on the books for very long.

That, I guess, is maybe how you would compare our models. I'd love to talk about the accretive issuance part because I think that is maybe really underappreciated. People look at the premium, right, and they view it as something that is bad. I think that's missing the picture. Yes, the premium adds extra risk, because if the premium goes down and it compresses from, say, 4x to 2x, that is a 50% loss in market price.

It can also expand if demand increases, right, or if, say, the market is pricing in something that is not within the NAV, because the NAV—you know how it works—is based on the last-round valuations. Sometimes there's a discount if, say, we have Series B shares as opposed to Series C shares, but it doesn't take into account if a company is going to do a next round very soon, or there's this major milestone or a huge amount of growth that is not encompassed in NAV. The public markets price these things on a more real-time basis.

But in terms of accretive issuance, a traditional venture fund drives its returns based on the underlying investments. If we're able to manage the premium well, we can have another engine for growth for the fund. It's what we talked about earlier for the private-equity/MicroStrategy model. They call it, in their world, multiple arbitrage—private-public-company arbitrage, earnings arbitrage. It's the same exact concept applied to venture capital.

Hypothetically, if we go through a bear market and there are no markups in our portfolio, and the value of the companies from a NAV or from a mark perspective stays the same for 6 months or a year, if there is a premium that is held—I'd say 3x, 5x, whatever—and we're issuing shares at that premium, that can increase the NAV per share. If we do it at a discount, that would decrease the NAV per share, on the other hand.

From a regulatory perspective, we're not allowed to do it at a discount, nor does it make sense for us to do it at a discount. Nor does it really make sense for us to do it at 1x or a very minimal premium, because we believe in the value of the portfolio, right? As the largest shareholder myself, my incentive is to maximize shareholder value. We're going to be very thoughtful in terms of how we raise capital and do share issuance in a way that is maximally accretive for shareholders.

I think that makes sense. Also, one point that I've always found interesting about this discussion, and then I want to talk about the specific companies as we wrap up, is that as these companies become—as MicroStrategy becomes bigger, or maybe not necessarily MicroStrategy, because Bitcoin's a large, liquid market—but any private holding vehicle that goes public, at some point it might be setting the price for the underlying assets, because these things aren't trading 24/7, right?

If people are willing to pay 2x for your NAV, maybe these companies are going to raise it to 3x in a year or 2 years anyway. People are just pulling forward that price appreciation. That might actually be the true price at that point, which is what I'm also trying to wrap my head around.

It's possible, right? Because I think if you think about Bitcoin NAV, the way MicroStrategy looks at it, they're valuing their NAV based on a liquid asset that's trading in real time, 24/7, and basically it's a mark-to-market asset.

8. Rapid Fire: The Portfolio

For us, we invest primarily in private-market assets that are not trading 24/7. If they went public, it is possible that they would trade at a different price. They could trade lower, they could trade the same, or they could trade higher.

As we wrap up here, I want to do a rapid-fire round of all the companies in your portfolio right now: the 30-second pitch for each one and why they're in your portfolio.

Okay. Figure AI, I think, is the Apple of this generation. If you just look at their videos and the robot that they've designed, I think it's really clear that they have taste and style that I haven't seen from any other robotics company out there. Definitely not the Chinese ones—not to knock them per se—but those look pretty industrial and cookie-cutter.

What's beautiful about Apple is that they made a product so thoughtfully, with design and user experience in mind, to a level that nobody else was doing. It's a beautiful experience having an Apple. That was what they focused on, I think. Figure has this very similar type of DNA, and you can see it in the content that they put out.

If I have something in my home, I want it to be awesome, right? I want it to be perfect, and I'm going to be inviting people into my home. I want to be proud of that. Having an iPhone, at some point, was a little bit of a status symbol, and I think you're going to have a little bit of that same thing happen with robotics as well. Obviously, I think they're going to be able to execute on that vision. They have the technical chops to do so.

Apptronik, I think, is—I would say—the next best, maybe a year or 2 behind Figure, playing in the private markets for humanoids. They were one of the earliest innovators of humanoids out there. They built the prototypes for some of the biggest humanoid companies that we know today. They have a lot of experience in hardware design and manufacturing, which some people might perceive as simple, but Elon has said that manufacturing for Tesla was harder than designing it.

That's underappreciated, right? They have a partnership with Stable, one of the few large-scale hardware manufacturers in America. They also have a partnership with Google DeepMind, which is one of the leading AI organizations in the world and has been working on robotics for more than a decade.

We have Standard Bots. We talked about them previously. I think that they're going to be a real industrial giant in America. They're going to power, basically, the reindustrialization of America. They're vertically integrated, which is really important.

I think that company is important to the American government and the American people, for that to succeed. We have Dyna Robotics. They are one of the best robot-learning research groups out there, especially in the field of post-training.

They were one of the first companies that were able to get these robot models to perform at a 99.9% success rate. That's what you need to be able to deploy these in the field. They can't be breaking; they can't be working only 95% of the time. It can't even be messing up 1% of the time. That's too much.

You can't have a human intervene every hour. The goal of these robots is that humans don't have to intervene ever. They're also building their own hardware platform, which is pretty differentiated as well.

Dexmo is taking an approach where they're working toward becoming the Unitree of America. I used to be skeptical about the Unitree approach because the hardware wasn't good enough to be used in a production setting. But it was really great in the sense that, for their business, they were able to sell thousands of robots last year, and they will sell many more this year. They were able to build the developer mode because it was really the only commercially available humanoid that people could buy conveniently and at a reasonable price.

There are a lot of research labs out there and, I guess, entertainment firms that want robots. If we think about what's needed in the robotics world, we need to do a lot of research, we need to do a lot of development, and we need to collect a lot of data. To do that, you can't do it on a screen; you need actual robots to do that.

Dexmo is the first company in the US to have commercially sold humanoid robots. They sold the first wheeled version last year, and they were selling to basically all of the major US research labs and some big companies. They're going to be launching a legged version pretty soon. I think that's later this summer.

What people, I think, don't understand is that robots are going to be a platform in the same way that the iPhone is a platform. People are going to be able to develop their own skills. You can think about it as a developer economy that's going to form, and that's what made smartphones so great, right?

They didn't just have all the experience from everybody at Apple; they were able to take it from everyone who was creative and smart in the whole world and make the iPhone more useful for everybody. That's what's going to happen with robotics as well. I'm going to have somebody create maybe a skill to have my robot be able to cook as well as Gordon Ramsay.

I mean, that would be pretty nice.

To do that, you need to have a robot that's commercially available for people to work on, right? Not everyone's going to develop their own robots, their own hardware, and their own manufacturing. That's really hard. You need somebody else to do that, right?

I think that's a really great business model that's underappreciated, and not many other companies are currently focused on it in America.

Well, maybe we need to do a part 2 to this podcast because I have an infinite number of questions to ask you, and I feel like we only scratched the surface. Unfortunately, that's all the time that we have for today, so, Andrew, I really appreciate you coming on. This has been an incredibly enlightening conversation. I know our listeners are going to love it.