Full Transcript

‍

Slava Rubin (00:00)

Welcome to the latest episode in our pre-IPO series. My name is Slava Rubin. I am one of the founders here at Vincent. You could check out our podcast like Smart Humans, also our daily newsletter covering all things alternative investments, the alternative investment report. We love doing these pre-IPO talks and sometimes we cover an entire industry. Today we're gonna be talking about humanoids.

and robots with us, the famous Jan-Erik from Sacra Jan-Erik, thank you for joining.

Jan-Erik Asplund (00:32)

you, Slava Great to be here.

Slava Rubin (00:34)

Absolutely. So on today's agenda, we're going to be covering this whole new wave. Obviously, everybody's talking about AI. They've been talking about it for a few years now. And I think that this next wave of discussion is going to be how AI and robots are coming together. And more often than not, we're going to be using this word humanoid, which is a brand new word for most people, but it's going to become pretty normalized.

⁓ We're gonna talk about what does this all mean, why now. We'll cover some of the main players that you should know about and also some of the emerging players. And then we'll discuss where this can all head. For what it's worth, timing is good. 60 Minutes just covered the same topic this past weekend. So we're right on trend with the zeitgeist. And as always, feel free to send in questions at any time. I will do my best.

to keep up with the questions and incorporate them into the actual discussion. For what it's worth, Jan-Erik just hear this. This is the first time as we opened up the session, we already have five questions that have come in. So people are excited, even before we even talked about anything. So let's jump in. First, word from our compliance department. and sorry, this session is actually brought to you with our partner, Augment.

Augment is an awesome pre-IPO platform, helps you to get access to all these private names, whether it's the tops in AI or defense or these humanoid companies, they're our partners, so we're excited to work with them. So nothing in this presentation should be construed as an offer to sell securities or a solicitation of an offer to buy securities. All investments involve risk and the possibility of loss.

including loss of principle and neither past performance nor forward-looking information is a guarantee of future results. We're about to dive in with our first topic, which is just some background. But since we already have a bunch of people on this, actually wanna start with a very simple survey, which is two questions and see how you all respond now and if your responses change at all.

as we end up completing. So question number one, how many of you are thinking you will invest into an Android, sorry, a humanoid company, a robotics company in 2026? Your options, which this year are either yes, no, or I'm not sure. And then your second question is how many of you think that you will actually own

a robot or humanoid by the year 2030.

We're gonna give everybody

All right, survey says, what do we got?

the results are in. So will you be investing in a robotics company this year? Look at that. We have 56 % saying yes, only 7%. So we have a very pro robotics. So they are all interested to Figure out who they are investing into. 37 % are unsure, looking to be swayed one way or another.

So if our production team is able to get the results of owning a robot, just feel free to throw it up when we get it. So, Jan-Erik why now? I've seen the movie Terminator 2. I grew up with the Jetsons. We've been talking about all this for a while.

We actually have other robotics companies like iRobot and others that have been struggling recently. So why all of a sudden are we gonna have humanoids and robots in our homes?

Jan-Erik Asplund (04:04)

Yeah, great question. RoboCop also, a really good movie in this vein. But yeah, there's kind of three big reasons why the last few years have brought all of this talk about humanoid robots. And so really it's kind of these three trends that all kind of converged sort of late 2022, 2023. And the first is the hardware side.

I think people really lump together AI and robotics, but the reality is that a huge part of robotics is, you know, the physical hardware that goes into the robots. know, that's not anywhere near a solved problem. But one of the big trends in sort of 2000s, 2010s was a lot of the hardware that goes into humanoid robots has become much more, much more commoditized and much more.

much more efficient, much more resilient. So one of the really big ones for sort of smooth motion is actuators. We've seen electric motors get very good. The sort of electronic control units have gotten really good propelled in large part by work being done on electric cars. So as electric cars have gotten much better, some of these underlying components have gotten a lot better. Same with microprocessors, cameras.

you know, cameras have gotten really cheap and then compute, know, sort of foreshadowing here, but there's a lot of GPUs available now for inference. So on the hardware side, it's just gotten a lot cheaper and more reliable to build, you know, a human-shaped robot that can move around and take action in space over time. And now, you know, the price has gotten down to...

1630K, it's heading down to the 5K, 10K range per robot. in China, they're really pushing the envelope on that side, on the cost. So we're getting cheaper than a car. The second big piece of it is AI. So this is where sort of 2022 comes out. Pre-2022, there really was no way to build...

an end-to-end AI for robots to be able to take in information visually and then decide upon some kind of plan to fulfill a goal and then execute it. And that has changed. So now we have the ability to give robots some data, have them extrapolate everything that's happening, and then give you some kind of output. Before, this would be a very slow process. So you could have robots do

pallets in an Amazon warehouse. But today you can encode these more challenging tasks using the same transformer model architecture that is behind chat GPT and things like that. So that has changed. There's still a sort of, it's not that better chat GPT means better robots because there's two different kinds of, there's the planning, which is sort of a symbolic language activity. And then there's the actual taking action, which requires a different

type of AI model, different sort of set of problems, which we'll talk about, but we've come a long way. And then the third sort of trend, maybe less driving the capability or feasibility and more driving the motivation is that, you know, we are facing kind of a massive cliff as far as demographics and manufacturing workforce. There's just not enough people to do all the work that is out there.

You know, the US alone is facing a 2 million person shortfall in manufacturing. So there's a great incentive on the part of industry to Figure this out. And a lot of these jobs are physically demanding, very repetitive, undesirable, dangerous. And what's more, you know, one of the key questions, and we could talk about this more, is why humans? If you can make a robot in any shape, isn't it just the ultimate narcissism to make it in human shape?

there's very good reason to make a human shape. And that's because every factory warehouse that exists today, you know, or 99 % of them are designed for the human form factor for humans who walk around and have arms around arm height and, you know, turn around and swivel. And you can do what Amazon did and build a robot. You know, for example, if you've seen the robots that go around on the floor at Amazon warehouses and bring stuff around very quickly, that's great.

if you have 12 to 18 months and loss of capital to retrofit a factory to work with that form factor. Whereas a robot that's a humanoid shape can in theory drop into any factory or warehouse, work alongside humans and one day replace them. So those are really the three kind of big trends to talk about in the space.

Slava Rubin (08:44)

Yeah, super interesting. ⁓ What do you think are the most obvious upfront use cases that we're going to be seeing in earlier days of these robots? How will they be used?

Jan-Erik Asplund (08:59)

Yeah, I think a lot of the earliest use cases will be ones that are not necessarily in the public eye. So, you know, probably not a humanoid robot necessarily in your home. The big ones from our discussions with folks and we've talked to some founders of these ⁓ companies, the big ones that they're looking at right now are ⁓ manufacturing and especially automotive. So Mercedes-Benz, BMW.

are already doing pilots here because ⁓ of assembly lines, very rigorous kind of regimented work, very physically demanding. So that's been a very big one. And then logistics and warehousing. navigating warehouses, loading, unloading, and also being able to have robots that work 24 hours a day, doing receiving in the morning, outbound shipping in the evening is really big. Smaller ones, oil and gas.

where you have very hazardous environments and you want to minimize human exposure. Some stuff in sort of restaurants, food service, I think military is one where we'll see it, where it will sort of appear first. That's where I think of these sectors. Yeah.

Slava Rubin (10:04)

So let me just push back a little bit on the need for a humanoid. So Tesla is already quite automated in terms of their kind of car manufacturing. And a lot of that is robotics and very little of any of that is actually let's call it humanoid robotics. So on both sides, why is, you mentioned this a little bit, but why are we so excited about the humanoid form factor?

versus not being excited already about all the other robotics momentum that's been happening? And what's the difference between the two? Or is it all one and the same and people are just excited about seeing like their own self image? And we as humans are narcissistic. So we like seeing like the human form factor more.

Jan-Erik Asplund (10:50)

That's an undeniable part of it, right? I think it captures the imagination. There's a reason that the great movies we mentioned are all involve humanoid robots and not just industrial square box robots. Also, I think a lot of the stuff that is impressive in robotics pre-humanoids like Amazon, for example, is internal to Amazon. You know, they didn't make that available to other companies because that's part of their secret sauce. But that also speaks to the fact that it had to be designed and engineered specifically for

Amazon warehouses and Tesla similarly has automation in their factories for welding and stamping and painting these very rigorous activities that take place on an assembly line or similar kind of environment. So they're a little bit in the shadows, know, behind the scenes. And then I think the thing with sort of these robots is that they perform these specific tasks very well. They're fixed in place and they're very specialized.

That's proven. So humanoid robotics much less proven. But the theory is basically that, you know, there is still lot of manual labor in Tesla factories. You have a lot of moving components around the assembly floor. You have a lot of quality inspections. You have a lot of more complex assembly work like wiring, interior components in cars that require more precise dexterity, right? Versus a robot designed to stamp a piece of metal into place.

So because of that precise dexterity, lot of these things like wiring harnesses together in seats, this has been impossible to automate until now. the other thing too is like I sort of alluded to, know, human noise can work anywhere. So Tesla and Amazon might be able to automate with these very specialized robots, but you know, your local factory, you know, which is going to shut down because there's not enough people to work in it.

do that, they would, know, but they could with humanoids bringing in sort of replacement labor that could slot into any factory anywhere. So that's sort of the promise of it. And I think part of why Tesla is building humanoid robots, even with their sort of extensive automation already in place.

Slava Rubin (12:54)

Interesting. Our producer mentioned that we have the second poll ready, so if you want to just throw that up. sweet.

Oh, 50-50. Nice. 50 % is actually pretty high for owning a robot by 2030. So again, it's a very optimistic audience here. If you ask my mother and her friends, it's going to be negative, negative zero for sure, even though they all want help with their laundry and dishwashing. So we'll see what happens. Speaking of which, you have this chart here, which to be read correctly,

This is how many human jobs are going to be replaced by humanoids, is that right?

Jan-Erik Asplund (13:30)

Yeah, this is a Morgan Stanley projection, it is basically the potential number.

Slava Rubin (13:38)

Right, so it's pretty low in 2028. It's at zero. And by 2030, it's only 40,000 jobs, if I'm reading correctly. But you fast forward 20 years later, and 20 years is a long way away, but then again, not that long away. Potentially 60 million jobs are replaced. And what do we have in the United States? Like 300 million humans or something, 350 million humans?

Well, no, I humans, Adults, I think like one third of those are not adults. So for simplicity's sake, I'm using really round numbers. We're talking about 200 million adults. And this is like 60 million. Again, I'm using really round numbers. We're talking about one out of every three jobs potentially replaced by a robot. Now, this doesn't say this is US only. it does, US jobs. So think about one out of every three jobs.

Human jobs replaced by a robot. That is fascinating. I'm still quite bullish. I think the humans will Figure out other amazing jobs to have. Probably better jobs, hopefully. But this is just a fascinating concept. Let me ask you, do you think that it's only the humanoid form factor that's going to win out come 2050? Or do you think we're moving towards Star Wars?

where there's gonna be all kinds of flavors of robots. Just to give you an example of that, right? So whatever, 150 years ago, there were no cars. Next thing you know, we Figure out how to do cars. Now, yes, there is kind of like the sedan, but next thing you have is the SUV and the van and the Sprinter and the truck and the 18-wheeler and the motorcycle and the unicycle and the blah, blah, blah, blah, blah, blah, blah. So what do you think?

Jan-Erik Asplund (15:20)

Yes, I think there will be definitely more than just the humanoid form factor. There's, for example, one company I'm forgetting the name of that does oil and gas kind of inspections on pipelines. And so if you think about that use case, it's one that humans can do, but it involves, you you have to get high into the air, you have to get into dangerous areas. And so it actually doesn't make sense to build that as a humanoid form factor.

because we can build something better. We can build a small robot that's agile, that can fly, that can climb, which is what we've seen. Boston Dynamics has a robot. If you're familiar with Spot, their robot dog, four legs thin. Our head of research, Marcelo, just texted me. It's called Gecko Robotics. Thanks, Marcelo. Great clutch.

Yeah, that's just one example where I think the use case demands a different form factor and we will see different form factors emerge. you know, another kind of gray area is that humanoid doesn't always necessarily mean 100 % human. So if you're familiar with, you know, the, the animal that's, that's half human, half, half horse, I think we'll probably see something like that in robotics. We were already seeing, you know, half human, human from the waist up and then a pedestal, you know, below or wheels below.

as sort of a hot, swappable base for a humanoid robot. So we'll see humans, we'll see humans that sort of moonlight as our transformers. We'll see completely different form factors as well, ⁓ flying, climbing, digging, that kind of thing for sure. And I think humanoids kind of capture the imagination, but many of these might be just as kind of significant in the end.

Slava Rubin (17:00)

So we did a poll and 50 % said they would have a robot by 2030, 50 % said they would not have a poll. This is just our audience. You're obviously doing a ton of research. From your research perspective, when do you think we're going to have actual robots being used, whether it's in the commercial setting or the personal setting? Are we talking next week, next month, next quarter, next year?

2030, when do you think like the first robots, know, kind like the early days of driverless cars going around town sort of thing.

Jan-Erik Asplund (17:41)

Gotcha, yeah, kind of like you can casually spot one. Yeah.

Slava Rubin (17:44)

It's like super novel. Yeah.

Jan-Erik Asplund (17:46)

It's a great question. And I feel like this, can't speak for the audience, but I feel like for me, this is what has me vacillating on, you know, the question of would you invest in them this year is that it's really hard to say. It's hard to have visibility into what is happening currently. But I would ultimately say I, personally feel like 2030, you know, I would expect 2020, 2030, 2031, now that we're in 2026, maybe I would, I would expect

to be chancing a glance at a pretty novel humanoid robot out in the world.

Slava Rubin (18:22)

I'm gonna push back on that. I think it's gonna be faster than that. That doesn't mean I expect my audience members to have their own robot so quickly, but I think that maybe even by end of this year, but definitely by 2028, I think there's already gonna be the early, early use cases of robots in the world. I'm talking super small amounts, but actually being used.

I think people will want to write a ton of pros and cons about that. Just like they did about driverless and it'll take a solid decade for it to become much more mainstream. And now we always have the way most of Tesla's and everybody's saying when is Uber going to have driverless cars for their entire fleet? mean, fast forward, that's going to take, you know, 20, 40, 20, 50, in my opinion, in the robot world. But I do think we're already see some of these robots. What's the what's the business model?

Jan-Erik Asplund (19:11)

Yeah, there's a few different ones that we've seen so far. You know, there's obviously the sort of buying it outright business model. So they, you know, build it for five, 10 Ks and sell it for more, which we've seen you can, you can buy humanoid robots from China already. So conceivably, some of the audience already has one in their home. The other big one that we've seen is the robot as a service model, right? Where, you know, borrowing from SAS, but you would go to an Apptronic or Figure or optimist if you're a factory.

And instead of paying upfront, you would just pay a monthly, yearly contract to sort of have access to that robot for that amount of time. Those are the two biggest business models we've seen.

Slava Rubin (19:50)

Got it. And so this can potentially be an entire separate talk, but there's the actual robots, which is the humanoids, but separate from the robots, there's all the potential new software that comes within this entire industry. Do you want to just mention that for a second, whether it's like physical intelligence, skill, feel AI or others, we don't have to do a deep dive here. It literally can be an entire separate talk, but if you just want to mention the difference.

Jan-Erik Asplund (20:15)

Yeah, exactly. Much like with the software business model, I think there's excitement about the potential for this kind of, you know, this attach rate, you know, kind of driving expansion by being able to dynamically, you know, add new capabilities to robots, which is a big part of these just becoming, you know, AI becoming so much more powerful in the robots, so much more capable and being such a big part of the sort of the brain, you know, of how they think. you can look at

I would imagine that we'll have what we have with Tesla, which is continuous learning, getting better, continuous software updates, data being collected from other robots, improving the fleet, buying new skills, that kind of thing. So yeah, definitely can imagine that.

Slava Rubin (20:53)

Okay, well let's dive into some of the players. Let's start off on the next slide with our first, which is Figure. Is it fair to say, well, let me not lead the witness. Who would you say are the leading players in robotics at this moment?

Jan-Erik Asplund (21:08)

Yeah, I mean, I would probably say Tesla, Figure and an Apptronik jump out. And just saying that, I know there might be people who are like, what about X? And the reality is that these three seem like the furthest along and most of them are prominent in terms of pilots and attention. But everyone's still pretty early. So I'll see that as a sort of disclaimer, but these three.

Slava Rubin (21:32)

Okay, so Tesla, Figure Apptronik, let's go through them and then we'll talk about the other emerging players as well. So give us some love on Figure.

Jan-Erik Asplund (21:39)

Great, yeah, so Figure founded in 2022 by serial entrepreneur, Brett Adcock and their pitch, they were one of the first big names here. The pitch is general purpose humanoid robot can work in many different environments and do human tasks. They started off as an early partner of OpenAI, which they sort of, they ended that partnership and they decided to go full stack designing everything from the hardware side to the AI models in-house.

really betting on kind of vertical integration and controlling the entire stack, both for cost and performance. And they've had some notable, you know, pilots, they've deployed robots at a BMW manufacturing plant in South Carolina. So they are doing some work over there. And yeah, that's really the headline. They have a few other partnerships going on with big manufacturers.

And they've talked about this robot sales and also robotics as a service business model going on. And then obviously sort of bearing the lead, valued at $39 billion. So ⁓ huge valuation, ⁓ very little de minimis revenue from what we understand. yeah, astronomical multiple on one of the leaders in the space. And yeah.

Slava Rubin (22:57)

I mean, it's nice to be Brett, right? Which is you say you want to start a new company, you'd start with your pre-seed at like $350 million post, God bless you. And then two rounds later, definitely have momentum, not trying to be negative here, but now have a $39 billion valuation off of not that much traction. Definitely no revenue, definitely not, you know.

We'll talk about the difference with Tesla and Apptronik in a second, but they are definitely considered one of the darlings. Lots of opinions. Some people think it's massively overvalued. Some people think it's massively undervalued because they are going to be the dominant robotics company in the world. lots of opinions on that one. So any other comments you want to say on Figure at the moment?

Jan-Erik Asplund (23:46)

Yeah, well, I think there's a there's an attitude among some, maybe the people who think it's undervalued. I think that this is incredibly capital intensive space to build out, you know, the thousands, hundreds of thousands of robots that are going to be needed. And there's a little bit of the 2020, 2021, you know, dynamic where, you know, anoint a winner and then ⁓ flood them with capital and give them the biggest war chest and they're going to So it's kind of like the soft bank.

you know, kind of soft bank thesis, which, you know, there's some, some, some value to that thinking in a, a, a cat, you know, in a very capital, expensive business. the flip side is if you believe, you know, that, that, ⁓ you can give a company too much capital, right. and make them, you know, a little bit lazier fatter than, other companies in the space. So, a lot of it kind of comes down to your, your perspective there.

Slava Rubin (24:37)

Great question from the audience. So the form factor of the humanoid is gonna have its own operating system. So there's gonna be a need for a different way of entering into that whole software layer. Figure has its own system, Apptronik works with Google, et cetera, et cetera. You have any perspectives on the operating system mode, et cetera?

Jan-Erik Asplund (24:57)

Yeah, it's very interesting right now because you're seeing very sort of, you know, varied, know, NVIDIA is one of the leaders here, Project Groot, and then Google DeepMind obviously is one of the leaders in the forefront. And both have been somewhat promiscuous sort of with their partnerships. Apptronik, as we'll talk about, has a sort of premium deal with Google. But so far what we've seen is a lot of the same

⁓ names across the software side. Figure stands out ⁓ for being kind of a full stack built in house. And on the other side, what surprised me a little bit is seeing, you know, basically everybody working with either, either Nvidia and or Google on the software side. So we haven't seen a lot of differentiation, which is interesting.

Slava Rubin (25:43)

And you got to give credit to the CEO founder, Iacoc, who, you know, some compare to Elon. So he's very accomplished. And I do think if it's a war of having to be able to raise more money, Brett is going to be able to accomplish that quite well. So I think that's part of the thinking. Let's go on to the next company, which is Apptronik. So Apptronik is interesting because it's really tightly connected to Google.

⁓ that has its pros and cons, but Yann-Erik, why don't you talk us through this?

Jan-Erik Asplund (26:12)

Yeah, interesting company, know, founded by former roboticists from NASA. They were really a focus as a company on building the actuators that go into robots for years and selling those actuators to NASA and other places. And it wasn't until the last five years or so that they really started to shift focus and building, know, of building a robot with a brain around those actuators designed for logistics manufacturing.

and warehouses. But as you said, headline point about them is having an exclusive partnership with Google, which is at the cutting edge of figuring out what these action models can do and how they can help robots take action in the world. Apptronik has a large emphasis on modularity, swappable components, like I've mentioned, legs and wheels to mounts.

and being able to configure a robot for different tasks without buying a whole new one. And that's sort of their key is believing that the path to scale and to making these robots 5, 10K is all about this kind of cost reduction around parts. And ⁓ as far as making it smart, I think they're betting on Google, which from history wouldn't be a terrible bet.

Slava Rubin (27:24)

just to double click on not being a terrible bet, Google just passed Apple, actually, as the second most valuable company in the world, after Google was kind of in the dumps just a little while ago because people were saying OpenAI was gonna eat Google's lunch in the AI world. I think Google has been doing a pretty good job of using all of its resources to deploy AI into its various distribution channels. And you can't really...

shy away from the fact that Google is doing quite well. I personally think this is just my opinion here, but I think the Apptronik partnership with Google is very valuable from this perspective. And I think that is a differentiator as long as it stays exclusive. So the valuation officially is 1.8 billion last, but the rumor is that it's 5 billion. I don't think you could really get into Apptronik below 5 billion these days. If you can, it's probably a good idea.

But it's funny because they have this Google partnership been around for a while and it's five billion dollars and Figure is 39 billion. Can you just again like explain that to me?

Jan-Erik Asplund (28:27)

I mean

Slava Rubin (28:29)

or not even explain it me. Tell me like your perspective on, know, why is it not obvious to just invest in Apptronik or is that a dumb way to look at it? You want to go after the thing that's really expensive.

Jan-Erik Asplund (28:41)

Yeah, I mean, it's a great question. Age-old question. think there's like, it'd be very little reason to have two companies valued at 40 billion at this stage. But the fact that there was one, you know, that's bigger getting all the attention and seemingly the traction, I guess, leads to this effect of investing in the winner continuously over and over. I don't think it's necessarily as easy as, you know, so invest in Apptronik. I think, for example, all these companies are so early on.

that if you saw what these robots are doing in factories, you'd probably be very underwhelmed. It's like they're standing in front of a car and inserting a pipe into a hole over and over and failing at it often. It's not impressive work. Everything is really early on the industrial side with what these companies are specifically doing. So it's all very uncertain. But Apptronik's definitely priced at a significant discount Figure, whether that's

valid. I'm not sure, but I agree that I think the Google partnership is really big, especially Google invented the transformer architecture. They've been at the front edge of every AI ⁓ sort of innovation in last couple of years. And their robotics models do some really interesting things. yeah, it's more interesting to me for sure than Figure.

Slava Rubin (29:54)

Where's Apple at?

Jan-Erik Asplund (29:55)

on AI or in terms of a...

Slava Rubin (29:58)

Well, Google has this exclusive partnership with Apptronik and we're talking about a new operating system from Figure. So my obvious thought is, where's Apple? Like, are they in this world yet or are they partnered with somebody?

Jan-Erik Asplund (30:11)

Yeah, I haven't seen, I mean, Apple kind of notoriously people probably know this, but you know, has had a lot of issues with it's sort of AI department, lot of big departures. Siri, you know, is notorious to terrible. not a lot of great stuff going on in the AI world at Apple specifically. They're reportedly they're starting to explore it. You know, as of this year, there was reports saying that they were, they were exploring humanoid robotics. think the latest I saw was they were planning to release this kind of

Pixar style lamp, which would sit on your desk. Not exactly a humanoid. I don't consider Apple really even a factor here. I think they're completely out of this race, it seems, unless something miraculous happens.

Slava Rubin (30:54)

Fascinating. Let's go to the next company.

Tesla, big player.

Jan-Erik Asplund (30:59)

Yeah, so Tesla is a biggie obviously. It's Elon betting that Tesla has gotten really good at manufacturing electric vehicles. A of those components that go into EVs go into humanoid robotics. And that if anyone can sort of hit the cost and scale targets necessary to make this happen, plus bringing the AI necessary to make them smart enough, it would be Elon and Tesla. Elon himself has said this will be the biggest product ever.

when it's done, which is classic Elon, you know, very measured statement. But the advantage they have is obviously manufacturing scale. know, they already have all these factories that produce millions of these very complex electromechanical systems every year. They got vertical integration on the supply chain. They know how to drive down costs and use off the shelf components to be cheap. And then you have sort of the AI stack where

the same GPUs that they've been investing in to win with XAI and Tesla before that are being used to power training for Optimus and can be used to power inference for Optimus. So there's this kind of AI side advantage. And the plan with Tesla Optimus is very similar to Figure in Apptronic. It's to deploy them on factories, on assembly lines, warehouses, and sell them to those external customers. So at least on the

Business model, there's very little variation in terms of how they plan to make money and who they plan to sell to. Some variation, for the most part, it's very similar.

Slava Rubin (32:30)

So I don't think this is science, but right now with Tesla's one and a half trillion dollar market cap, what would you say is the amount given to the robotics aspect? Again, this is very much art versus science in this conversation. But what do you think that premium is that they're getting for this?

Jan-Erik Asplund (32:54)

question. I'm not sure it's being priced in too much, know, explicitly right now because it's so early. But I do think that some of the things that are priced in like the work on batteries and AI will directly influence the success or not of optimists. So, yeah, it's it's really tough to say. but I do think it's definitely a part of the sort of, you know, exuberance about Tesla.

which is so frequently doesn't look like it's based on an earnings multiple evaluation. So I think it's there maybe 5, 10 % if I had to give a number.

Slava Rubin (33:32)

So not trying to sound like an Elon fanboy, but why not think that with all of the manufacturing experience, with all the car technology experience, with all the AI experience, why not think that this is gonna be a runaway win for Tesla?

Jan-Erik Asplund (33:51)

Yeah, it's a question. You know, it's similar to thinking, you know, it's kind of like the previous point about why not think that Figure will win with the most money. know, Tesla has also a lot of money, know, virtually unlimited money if you are pulling from Elon and they have all this experience. You know, one answer is basically that everyone is almost still at the starting line when it comes to building humanoid robots and the thing that actually moves them sort of forward

that fundamentally drives progress is your access to good real-world training data off of factory floors and other deployment environments that you use to then make your model better and better. And you'll see it in commentary that Tesla has all this data that they've collected. Well, it's only off of cars and it's only of roads. So road data doesn't really help with a humanoid robot navigating a...

you know, a factory floor. If they had that, then they would pretty much be winning. They would have won because they would have all the data that they needed to build the best robot, the smartest robot in a humanoid. So even Tesla is roughly in the same boat, you know, give or take as Apptronik and Figure. The robots have to collect data as they go. And that's limited by your ability to manufacture them, get them into real environments and have them complete tasks.

and then have humans intervene to fix it if they do it wrong. And it's not something you can just snap your fingers and have millions of robots doing these tasks, obviously. this is sort of the, you this is what you say if you are a founder not in the top three humanoid robotics companies, right? That Tesla does not have the data, no one has the data and having that data is really the key thing that will drive the flywheel that will make these robots.

able to sort of satisfy what they say on the tin.

Slava Rubin (35:37)

Is that the bottleneck? Is that the main bottleneck? It's not.

Jan-Erik Asplund (35:40)

From what we've seen. Yeah.

Slava Rubin (35:43)

It's getting the data to make the robot smarter in terms of how they physically interact with the world. Is that right?

Jan-Erik Asplund (35:48)

Yeah, exactly. There's no, know, chat GPT was trained on the entire internet. ⁓ there is no such sort of equivalent corpus of, of information for robots, but the AI is really good. The hardware is already really cheap. You can already buy a 5k robot from China. so it really is. Yeah. The training data from what we've seen.

Slava Rubin (36:07)

I love that you just brought up China, because I think that's the elephant in the room. think, again, this is my opinion, but I think China is super into humanoid robots because they need to be. I think they have a population issue with their single child policies that's catching up with them. And I think they have an economy issue. And I think they're going all in with robots. What do you think about?

the Chinese companies and what is this gonna look like 10 years from now, Chinese companies, Chinese robots versus let's call it US robots or other robots. We haven't really, from a car manufacturing perspective, taken on a lot of Chinese cars in the United States, right? We're really much more domestic, but do you see that being different or the same with humanoids?

Jan-Erik Asplund (36:58)

Yeah, yeah, great points. know, China has huge advantages on manufacturing. You know, if you're familiar with like Shenzhen and Shein and Foxconn, you know, you'll know that the supply chains in China are really, really tight, really well integrated into these areas, much better than the US. And they have, you as you sort of pointed out, a population problem. Plus right now, you know, they have a very dedicated workforce.

you know, to sort of supply the work for these factories and really good pricing on all the components that go in. And so, yeah, right now, the split in terms of components, like who makes components for human robots is roughly 50 percent China, 50 percent everyone else. And they are ahead so far on shipments. So, you know, that's kind of the picture. If you are, you know, concerned about that, it looks very bad. I think on the other side,

There's several factors that help complicate the story a little bit. So one is, like you've said, we don't have a lot of Chinese electric vehicles in America.

Slava Rubin (38:02)

But we do have a lot of TVs or a lot of other electronics.

Jan-Erik Asplund (38:06)

That's right. That's right. I think what we have seen in our research is that these political barriers of, you know, importing are expected to be, are expected to be an impedance for, for humanoid robots because of how much of it is driven by defense and industrial applications. So, so especially there, it's going to be harder, most likely to import robots from China. It'll be different, you know, sort of, it's different for like sort of consumer electronics and

and home electronics, that kind of thing. I think it's going to be faster.

Slava Rubin (38:38)

I mean, because with Nvidia, for example, you know, the administration is trying to stop from sending some of the top AI chips, right? And I can see similar and different around the debates and arguments that will happen around who can send which humanoids to who, and who's allowed to get access to what. I think it's going to become a fascinating conversation five, 10 years from now. You mentioned this and we have many questions actually in the Q &A about this.

What do you think about the militarization of these humanoids?

Jan-Erik Asplund (39:11)

Yeah, it's really interesting. Speaking with some folks, Figure, believe, is the one that has sort of veered away from defense. I think there's always some trepidation about defense among robotics, people interested in robotics. What I will say is, obviously, Andriil has talked about robotics for sure, mostly in terms of

augmentation of humans. So imagine you have more powerful legs that allow you to run for a much longer period of time. I think Palmer Luckey is more bearish on humanoid forms exactly replacing human soldiers. But the other part that we've spoken with folks about is the use of humanoids in defense might not look quite like you're imagining if you're imagining sort of a Robocop ⁓

a deployed Robocop, it's more of like a, it might be much more in a maintenance logistics capacity, at least at first. And then over time, it might get to the point where, you know, ⁓ a humanoid robot is responsible for handling, you know, a piece of equipment in the field. Let's say a mortar, you know, which is dangerous, loud and potentially exposed to enemy fire. And so instead of having a human, you know, go over to check on it.

you would have a humanoid robot potentially do that. Potentially you'd have a humanoid robot driving a Humvee, so that instead of a human doing that, it's a robot, which is ironic, not having an autonomous vehicle, but we have millions of Humvees. So it's really the driver that needs to be automated. So I think in defense, what we've seen from so far is an emphasis on these kinds of auxiliary support, logistics roles that

you know, depending on your view on it, might be cool, you know, it's less about sort of creating an autonomous killing machine and more about helping people stay safe out there. So that's what we've seen anyway.

Slava Rubin (41:00)

So there's a bunch of other players. Can we go through some of the next few slides where you comment quickly on some of the other players?

Jan-Erik Asplund (41:06)

Yeah, 1x. If you haven't seen their YouTube video of Neo, their robot, look it up. It's a household ⁓ humanoid bipedal, vaguely friendly looking, washes the dishes, does the laundry. And the demo is groundbreaking stuff. The Wall Street Journal also covered it. How far along it actually is, how able it is to do all this stuff programmatically, remains to be seen.

They have pilot units out in assisted living facilities. I believe you can pre-order one in the US unless you have small children in your house. if you want to have one in your house soon, then that's probably the way to go.

Slava Rubin (41:46)

And they recently raised that at 10 billion valuations, so no joke. And are they the ones that are remote operated?

Jan-Erik Asplund (41:52)

That's a good question. Everyone has some aspect of teleoperation because it's just key to keeping these things actually operative in the field. so they don't mess up right now as they're still acquiring training data. Theirs did look particularly teleoperated, I will say, similar to Tesla's. So probably.

Slava Rubin (42:12)

And when I say teleoperated, meaning like a remote control car where somebody is actually controlling it from afar, which makes it a bit complicated in terms of compliance or privacy, et cetera, as to who's actually seeing what at home. So definitely a factor as these robot companies get built out. What's the next one?

Go for Boston Dynamics.

Jan-Erik Asplund (42:32)

Yeah, Boston Dynamics is really an OG, know, founded in 1992, spun off from MIT. They've had demo videos of their robots, you know, jumping, dancing, doing very cool things for a long time, acquired by Hyundai in 2021 for a billion. They're also partnering with Google. They're one of the more kind of long, you know, standing names in the space, and they have, you know, a little more traction on commercialization. They have these things.

Like I mentioned, they have the dogs, you know, working in oil and gas. And I think they'll definitely be a really good bellwether to see, you know, kind of where the whole space is. Because, you know, they've done a lot of really the cutting edge work here. But, you know, the other angle or the other take on them is they've been like five years away from, you know, really commercializing it for the last 15 years. And so it's not really clear if they can compete with like.

Tesla and Aptronaut configure when they get up to speed.

Slava Rubin (43:25)

Right, but what I do like about what they're doing and really the opportunity is the industrial focus, the manufacturing industrial kind of shying away from like the cute human kind of consumer play. I do think kind of manufacturing industrial will have a lot of use cases for the robots. Next one. Let's go to the next slide with some of the other emerging players. Go for it.

Jan-Erik Asplund (43:48)

Yeah, so Agility is probably the most mature here. They've raised over 400 million. They have a factory in Oregon. They're really focused on the robotics as a service kind of business model use case and looking to, yeah, they deploy some of these at Amazon fulfillment centers and other warehouses. Unitree is sort of the Chinese wild card.

They're known for building very cheap robots. they're one of the biggest names in the Chinese humanoid robotics world, a beneficiary of a lot of money from the Chinese government and focused on really undercutting Western competitors on price, scaling fast and dominating the market. They're filing for an IPO in China. And then Persona, founded by another sort of ex-NASA guy, they've raised about

27 million building for industrial environments, a lot of these declining jobs that are too dangerous or undesirable for humans. And yeah, very targeting, very niche. So oil rigs, chemical plants, nuclear facilities where that willingness to pay is very high.

Slava Rubin (44:54)

And these are all very different stages. So Persona pre-seed, something like $800 million, I think. You have agility, last valued at just over $2 billion, has definitely some momentum. And then Unitree already, this is Chinese company, like you mentioned, looking to IPO soon. And if you go online, you can see quite hilarious videos of Unitree robots dancing with entertainers and

you that's kind of the entry point. First you entertain and then after that do like these really hardcore tasks. All right, let's go to the next slide, which is really where can this all head? I I'm surprised how long we talk, but there's so much interesting stuff to cover. What is this look like in terms of the base case as to what our future looks like and what's the bull and bear? Not so much just about the numbers, but what are the factors that go into this?

Jan-Erik Asplund (45:45)

Yeah, so I think the base case is something like, you know, might say the base case is kind of like AI continues to improve at the current rates, roughly, and hardware continues to improve at the current rates. And I think that that is a really good outcome for humanoid robots. You they probably become good enough for actual deployment, non-pilot by around that 2030, 2031, you know.

period that we looked at and cost comes down so it becomes more affordable to adopt it. And I think again, this is primarily manufacturing and logistics. So I think the main things that sort of move that up and down, in the bull case, it's like foundation models get a lot better. They don't continue to improve at this current rate, maybe they improve faster, right?

achieve kind of more true generalizations, being able to learn new tasks without so much training. So without all that training data and adapt to novel environments that they haven't seen before, novel tasks, you and if that happens then, and unit costs come down, that might not even be necessary, but you know, we would see much faster adoption.

much quicker getting to that 30 % of US jobs being automated by robots. You might see consumer adoption, which just to mention, one of the reasons that these folks ⁓ make friendly looking robots for industrial use cases is because they all want to eventually one day put a robot in your house. So that's part of it. And ⁓ in the bear case, I think the main thing is kind of the opposite, right? The AI progress plateaus.

scaling stops scaling and prices of robots stay roughly where they are. And we don't really see these economies of scale. We don't see the pilots do super well. And maybe there's other sort of wild card issues, safety. A lot of the worries that people like Palmer Lucky have about robots are around them working with humans and having

I don't know, thousands of pounds of force that they're able to deploy with an arm or a leg. And maybe the AI gets confused ⁓ for a moment and they punch you through a wall, right? There's a safety problem. There's political regulation problems that could come into play. But the biggest one I think is that the AI stops sort of improving at the same rate and we don't get the training data needed to make them really functional on a wide scale.

Slava Rubin (48:11)

Awesome. We've spoken here for quite a while, almost an hour. We could talk quite easily for another hour or two. I'm going to ask for a producer bring back our only the first question is who's thinking to invest into a robotics company? And in the meantime, while people are voting on this for 20 seconds, Jan-Erik, how would you, I'm asking you personally, you Jan-Erik, are you thinking to invest into robotics this year? And if yes, which one?

or two do you like?

Jan-Erik Asplund (48:42)

Yeah, probably not. I do think I like Apptronik as a company a little more. I like the story, know, former NASA guys making ⁓ actuators in Austin and then realizing that they could build a human robot. And I really like, you know, them working with Google, who I think have really amazing AI models around this. But yeah, that would be my personal ⁓ preference.

Slava Rubin (49:06)

Nice, and what are our results here?

Jan-Erik Asplund (49:09)

I hope everyone's not no. Okay, well it's about the same.

Slava Rubin (49:12)

It's about the same, about the same. Great. So my opinion is I'm quite optimistic on this trend. I'm not optimistic that the results will happen quickly. So I think you need to close your eyes and ask yourself, which of these companies will still be around and be doing good work in 10 years?

And if you don't feel comfortable saying, believe in these guys 10 years out, then I'm not sure if you should invest in them. That said, again, not trying to be an Elon fanboy, but I do think Tesla will be around 10 years from now. And I think that Tesla will have a strong robotics capability. So I think that's a very interesting investment opportunity. I don't think you can make.

a ton of money investing in Tesla because it's super expensive. So if you were thinking to take a little bit of a flyer and pick one of these other players, I think that's interesting as well. I don't really have a ton of perspective as to which one's better than another, but I do think that from a value opportunity, seems like Apptronik is an interesting play. Figure has a lot of opportunity to become a open AI or an anthropic type player.

So all could be very interesting. And a lot of these others that we covered are a bit younger, but there's no reason why they can't become worth 10, 20 or a hundred billion dollars, which is a massive result. so thank you everybody for joining. Thank you, Augment for being our partner where all of you can access these types of companies if you're interested in buying any of them. you know, look forward to talking to you about our next topics and today's robotics discussion is over.

Have a great rest of your day.

‍