Robotics Could Be the Next AI Infrastructure Trade - Here’s How We’re Playing It

Simple math implies that there's a 19 to39 billion just flowing into the semiconductor demand from the robots. Now, this doesn't even include like the chips used to train the models or the networking and storage infrastructure. >> What's up everybody? It's LG set here and welcome to Milk Road Stocks, the daily market show that can't wait to have my hands harvested for humanoid use after I'm killed in the great robot war of 2028. Today's September 7th, 2026, recording way back on September 2nd. Listen, the robotics wave is coming. For investors, it's a matter of when, not if. And today, our leading portfolio manager, Melvin, is going to dive into that humanoid wave and why Goldman Sachs just revisited its robot forecast by an added 600% by 2035. Yes, they think there will be six times more robots in use by then than originally thought. Plus, Melvin's also going to tell us where to look and how to spot this trend as it slowly takes shape. Melvin is a smart guy. He's our portfolio leader for a reason. He nailed the bottleneck trade this past year buying Micron, Nebius, and AMD way before they ran. So, if you want more of his calls and to kind of track robotics as he explores it and you want to see his full holdings, all of that is in Milkroad Pro, which is just a dollar at the link below. Melvin, I haven't heard enough about robotics this month. And I'm dead serious, man. Like, I think this is a huge blind spot for everybody, mainly because we don't know when or what's going to happen, but we know that it will happen in some way. Uh, so I'm excited to learn about robotics with you today. >> Thanks for having me back, um, LG. And, uh, you haven't been watching the robot Olympics that happened recently. >> I talked about those with Scott Walter when I interviewed him for the show uh, at all those like, you know, you know what he told me? I'll give you a little anecdote, and he said this on the show, it's worth giving back to you. He's like, the stuff that they're showing you in the robot Olympics are not their newest models. That's just to make you think. If anything, he's like, "That's like a scop by China to make you think that their robots suck." He's like, "That's that's just distracting you." >> I'm sure some 13-year-old kid made those robots in China that they had to deal with for school projects. Like, you know, >> that's what was in that's what was in those videos. So, for anybody who doesn't know, we're referencing to all the the all the amazing videos that you probably see on Instagram or your socials or or Twitter or wherever that just show these like robot Olympics in China where you have robots running around a track. Uh, and and at the end of the running, they literally slam into a wall and then like explode. Uh, and they're incredible videos, but like you're saying, it's like in one of the videos, Melvin, you see this like young guy walk over to the robot with his like giant remote control. This is clearly this guy's like homemade project. So it's like I think the internet is thinking that this is the latest advent of robotics and it's and it's nowhere near that, you know, like this is just this is high school high school project level >> because that's just China, right? We haven't heard much about Elon and what he's doing with Optimus in a really long time, which I think is kind of the strategic play from Elon. He just doesn't want to release what he's working on and all that. So there's definitely like a Chad GBT moment coming for robotics, you know, somewhere in the next few years. We don't know when, but there is a lot of lot of projections that came out recently and um I wanted to go go over some of those with you guys and talk about some of the stocks that you know we're going to we're going to you can potentially invest in. So, >> excellent, man. Well, I'm stoked. Let's get into it. >> Yeah. So, the first chart, uh Goldman Sachs just made one of the largest robotic forecast revisions, and I think it shows you how quickly robotics as a industry is changing. Um, Goldman previously expected 1.4 million robots to ship in 2025, but now it expects 6.5 million. So that is about roughly about 5x times its previous forecast. And what's interesting is Goldman also believes the market could grow from approximately 3 billion in 2026 to 27 billion in 2030 and 138 billion by 2035. I I personally think they're vastly underestimating the the numbers here. I think this is like a much going much much higher because if you look at Morgan Stanley's projections, they're they're even crazier um because they're they're more bullish over the long term. If you go to the next chart, it shows that uh the world potentially reaching 1 billion human robot robots and approximately 7.5 trillion in annual humanoid revenue by 2050. So we have like two different projections right now. And that that's that's where we're at in robotics because nobody really knows how big this is going to get. Everyone just knows this is coming. But as we get more ro as more robots get deployed into world um and um you know starts to bring in results I think these numbers are just going to get adjusted higher um so that's why I remain extremely bullish on robots. >> Mhm. I like that. I like the number you shared on the last one because basically Morgan or Goldman Sachs is basically revising their humanoid forecast by 6x like a 600% increase and that's from their prior forecast. When did these forecasts come out, Melvin? Like is this like every month? They do this just once a year. >> Yeah, they do it every like 3, four months. Like I've seen this chart before. >> Okay, so something has happened in the last 3 to 12 months, let's say, for Goldman to say like, listen, we're going to have six times more robots than we thought in 10 years. >> Correct? >> Right. Something has happened. >> Something's big is happening. And I know these numbers, you know, like Morgan Stally numbers can sound a little insane, but but the opportunity like, you know, extends far beyond those companies building the robots. Um, like every humanoid robots, if you think about it, needs AI models, semiconductors memory sensors actuators, batteries. You the list goes on and on. And these robots will also require, you know, um software to manage fleets and monitor safety. Um so what I'm trying to say is this is a huge huge market. It's not just going to be robots like like Tesla or like Figure AI. That's going to be the winner. The winners are going to be who are supplying the supplies to these to these robotic companies. So the point of this you know uh this video is I want to explain why the forecasts are rising, what has changed in the technology where robots are already doing useful work and which layers of the supply chain could make the most um uh money. So that's what this this video is all about. >> Absolutely. And we did and we did a video like this a couple months ago, too. And I think it's good to revisit this and and you've got some new info here as well. And the other thing I just got to add a bit of how I see these charts as well, Megan, because I think I think you you nailed it that it's like people look at this and they're like, where are you getting this? Like why why do you need to say this? Like Goldman Sachs, are you buying robot companies, you know, just buying more Tesla? And it's also like, you know what this where this kind of stuff shows up is in like venture, right? Where people that want to start robot companies or want to get more funding, all that kind of stuff. They need this type of chart to show people to be like, give us some money so we can go actually build this. And it's also something that you know talking to someone like Scott Walter obviously from Robo Strategy it's very clear for them as they they kind of collect um a lot of different private companies into their portfolio. Uh and also you know as everybody paints a picture of of uh you know the US potentially falling pretty far behind right so you need this kind of stuff to motivate money to get into this space. Yeah, and that's exactly the point. And actually, in fact, there was uh last quarter, I believe, robotics investments from BC deals hit an all-time high. Um I think it was around $2 billionish um back in 2021. And if you look at quarter 1, 2026, now we're looking 16 to 20 billion worth of investments. you know that you know 5x essentially in the last um um in in the last like four or five years but we're just getting started money is going to flow the same way the money flew into AI in the last four or five years because if you look at it the capex didn't really accelerate till 2025 when the results starts coming you know started to come in now we see the capex going from essentially from like 60 to 70 billion in 2023 to 2024 for somewhere around that number to all the way to 700 and it's going to 1.4 trillion and that is what's going to happen to robots eventually. >> Yeah, exactly. Okay, let's get into it then. >> Yeah. So, why are the uh forecasts rising so quickly, right? That's cuz the the biggest reason is several technologies have limited um you know uh previously limited human humano robots from improving at the same time. Now we know AI models are becoming better at understanding instructions and planning actions. We have agents and whatnot. Um onboard you know processors are becoming more powerful like while cameras and sensors and batteries and actuators are becoming smaller and more capable. Um now if you look at stimulation platforms like the one Nvidia has you know that can be used to train millions of robots you know without physically without robots having to be physically in a warehouse to you know collect actual data they can stimulate all this and also if you look at companies like figure AI or like u Tesla um they're also collecting more data real world data inside their inside their factories um they're being tested like uh in warehouses, distribution centers where they're actually operating in like you know operating conditions. So every like success successful movement or like every failure gives them more more like information that can be used to improve the software's software hardware reliability. This is like the same playbook we see in Tesla FSD, right? that I believe Tesla has like 8 billion um worth of miles in FSD um and those are actually making their models and they're making their vehicles better and which is what we're seeing in um in the robotics. But what's interesting is physical AI right becomes completely different from you know AI we use currently a language models like Chad GBT for the most part only needs to produce like models and results and whatnot but if you look at robots right it must produce an action it has to identify an object understand the task decide how to approach it estimate the weight shape move the joints correctly there's a lot of moving parts you know that's uh that's associated with robots but also at the same time you know consequences right of these you know mistakes for robots are much more serious. So if a chatbot produces an answer wrong you can like simply re generate it or like have the agent run again. But if a robot applies too much force in the you know walks when when it's walking or it can damage someone you know or injure someone >> feeds you too aggressively or something like that. >> Exactly. you know, hug hugs you too tight, kills you. >> Exactly. And they have like robots in China training kung fu and stuff, you know, like imagine getting kicked by one of those. You're like absolutely like you're going to get you're going to be dead. Okay, >> just going to pause there for a second to point out that the market is showing signs of something kind of different happening. And our analysts at Milkro Pro are all over it. They spent the last couple weeks making a lot of trades, getting out of some positions, and then getting into a lot of new ones, getting ready for the next wave of robotics, space, or even kind of picking some different AI winners. If you want to see what they have in their portfolios, what position is their opening, it's just a dollar in Milkroad Pro at the link below. What are they made of? What's What's the shell made out of? Like, what are the hands made of? Titanium. >> Probably vibranium. >> Yeah, really. Yeah, exactly. What's Yeah, I want to know. I want to know. Same with the ebikes where they have to c they have to cap the speed that you can go and you have to go to like a sketchy guy to unlock it. For robots, it's like I want the robot that can't apply more than like 5 lbs of pressure. Like I don't want anything more than that, you know? And I want it to be capped. I want it to be capped. >> Exactly. So that's why like physical AI is like much harder than digital AI because but that also makes it more valuable, you know, cuz if you look at robots, they have existed for decades. I I think in the in the the bot bot strategy guy video he said it started first robot was like in 1960s or something which is which is insane to even think about. Um but but the real question is what makes humanoid robots like different right traditional robots like industrial robots are extremely simple and effective at like simple and repetitive tasks. A robot arm for example can weld the same section of vocals thousands of times at incredible speed precision like same like that's what's inside Tesla factories right the problem is that these systems are normally programmed for one narrow task and installed at very controlled area you know nothing really changes like if the product changes or the parts move the company needs like company needs the machine to perform a different task the entire you know, system may need to be like reprogrammed or physically redesigned. This can take time and like require a lot of specialized engineers, cost a lot of money, but humanoid robots are supposed to be like much more flexible. Instead of being designed for one specific task, they're built to learn several different tasks and move between them. Um like a robot can like I don't know like unload boxes during one part of the day. Um move diff materials during another um eventually assist with uh basic assembly or inspection. What's important is the human shape is also important right like factories, warehouses tools shelves doorways they're all essentially designed for the human body, right? like car um a humanoid robot can theoretically work inside these like existing environments without forcing the company to rebuild the entire facility around the machine. Now this does not mean like human body is the most efficient design for every job like for example if we like a robot in wheels right can will usually like move faster um consume less energy than a robot walking on two legs. um a fixed robotic robotic arm can also perform like repetitive tasks faster and more accurately. So I'm not saying that I mean this is this is going to work but I'm saying like a human or robot is much more flexible can do a lot more than just a you know previous robot that that was in factories that that is designed for one specific ch job and that flex what's interesting is that flexibility that a humanoid has is could change economics of automation. A company like may not want to spend millions of dollars like rebuilding their entire workflow and the production line um to automate one small task. But what's interesting is it may be willing to purchase like or rent a human robot that can enter existing facilities and use the same tool as workers and move between several job you know. So that's what makes this all interesting. Um and and what the figure if you look at figure which is a which is another AI company I would say probably the second yeah there's a picture of that um they're probably the number two in the US uh behind Tesla Figure is already testing his humano robot inside you know BMW manufacturing facilities these robots are trained to handle like very very repetitive tasks such as like moving sheet metal components positioning for uh supply and you know supplies and whatnot. Um and then you have agility robotics you know if you go to the next slide is also deploying it digit robots um inside yep inside lo their logistic facilities um you know this is a bit weird looking but it's fine uh where they can like move containers and materials and you know all sorts of different tasks. Um there's a bunch more companies doing this. I'm not going to get into all that. Uh the point I'm trying to say is you know these are getting deployed in factories and um that get that the and whoever deploys the most amount of robotics inside these companies is going to be the most valuable because of real world operating data. So a robot can perform perfectly in like a control demonstration or whatever it is. But when it encounters like a you know I don't know a damaged uh like a packaging or like poor lighting or like an object being in the wrong location that robot wouldn't know what to do unless you have the real world life problems. I mean there's only so many so many tasks you can simulate in you know in a in a fake world scenario. These things have to come together naturally for it for it to improve. So, >> you know, the infrastructure question is a really interesting one to me, Melvin, because it it's it's almost funny that it's like, well, we have to build robots that fit our world. >> Yep. >> Like, why? You know what I mean? Like, it's it was like it's such it's like it's like an infraplay, but it's not. It's like, well, we have to make them fit what we have, right? And it's like because we can't change we can't change houses. We I'm sure factories over time will probably look significantly different because you don't need why make them for humanoid robots when you can just have a bunch of hands all over the place. Like I'm sure there will be one day like just a a robot only factory where you don't need space to really walk or something like that, right? But I think it's funny that it's like well we got to make them that the only use cases we really think of is like factory workers and uh home assistants. You know what I mean? There's no other things that we can think of and as a result it's like well we'll have all this cool new tech, but we'll still be stuck with the same old homes we had for a long time. I just think that that's hilarious versus like inventing cars. It was like okay no no get rid of these dirt paths for freaking horses and should get rid of these wooden wheels and crap. It's like, no, no, we need you need concrete and you need to make roads, right? Uh but then as a result, it's like, well, have self-driving cars. Why do they have to be fourperson vehicles, right? It's like, it should just be a little pod that drives you around. It doesn't need a steering wheel. It should just be a little box you sit in. It drives you around. We don't need these giant highways, but it's like what are we going to do? Tear them up? You know, it's like they have to be there. So, it's just funny. We're we're you know despite all these amazing advances like we're stuck in this kind of like in this old this way of building the world from like the you know post-war era that now we're technologies getting beyond that but we're stuck with the infra that we have right >> yeah but there will be a point in time where robots and AI get so good to the point where they'll start to do all the building and we don't even have to worry about it. Maybe they'll have their own houses that you know they can specially design and build. >> Yeah. Robot houses or something. I don't know. I have a little robot house out back like a dog house and at the night at night my figure robot goes in there and folds himself up and just sleeps in the in the robot house. >> Maybe like it could be well like you know Iron Man like you know they he has a fleet of robots that you know designs robots and makes robots and like you know assembles them, repairs them. That's probably the world where we're headed to. We just have a house that full of robots and it'll you know run it run as its own. >> Mhm. Mhm. Absolutely. And talking to Scott Walter last week too, it you know, and we were kind of mentioned this earlier too with like apprenticeship, learning hard labor skills, uh or even things that require like kind of an odd time of day, like you know, around where I live, it's like, oh, a power line comes down in the middle of the night. It's like people have to be called out of their beds, you know, and it's going to take a long time and those people are going to get paid a lot of money because they're working overtime and all this different is going to up their whole day. or it's just like okay well there's a crew of robots that is like the you know you know uh the power people they come from the power company they fix things it's way more efficient it's not going to happen tomorrow but you know once we have that critical mass of robots who have 6 million 10 million or whatever in 15 years um I can envision that being a really really helpful use case for governments and companies right so >> exactly or or imagine imagine in a world of natural disasters or something I just had a big tornado right that happened and the trees fell down everywhere the roads were blocked everywhere where imagine you could deploy like 10 20 robots and just move the trees or something point, >> you know, >> disaster relief to get into Yeah. get into places that are too hard to get to and bring people supplies and help them and dig dig people out of houses of collapse and stuff like that, right? That's a really good point. >> We saw the Nepal flood. Imagine if you can deploy >> thousands of robots and they're looking for people and they're rebuilding and that that's probably where we're headed. >> They don't get tired, you know. They don't get tired. They don't need to eat. They're super strong. Um Yeah, >> if we lose a robot, we lose a robot, you know. So, like, you know, >> and they can live stream so you can watch >> that, too. That too. >> You can pay. That's how they're going to make money. Mel, we're going You could pay to subscribe to these. People love disasters. I'm sorry. It's a human tragedy, but people people are addicted to watch. >> I'm sure they'll find a some way to bet on these robots. I'm sure that's what's going to happen. >> Now, now you're talking. This is my world. Okay, let's continue. We're way way off track. >> Now, we got to look at these numbers. Show me these numbers. >> Yes. The second major coh change is the cost um curve. Um so Goldman estimates that average human selling price um could decline from approximately 42,000ish to in 2025 um to around 21,000 by 2035. Um the estimated bill of materials which is the cost of like all the components inside the robot could decline from 28,000 um to roughly 13,500. Um, at the same time, like I mentioned before, annual shipments are expected to grow from fewer than 14,000 15,000 robots in 2025 to approximately 6.5 million in 2035. Now, this is extremely important because a thou $100,000 pro prototype that barely works is um like it's cool. It's a cool technology, but it's not it's not a mass mass market product. a robot costing like 20,000 or 30,000 that can operate like reliably reliably ex across like multiple like shifts create like a whole different dynamic for these companies. There are cheaper robots like there was a robot that got released that was like $300 um you know or there was like Uni makes robots like $3 to $4,000. So there are companies that are you know making making robots like cheaper but they're they're essentially useless now. they can do like kung fu and all that all that kind of fun stuff, but in practicality they're not they're not they're not really useful. But another important um important insight insight inside these numbers is that Goldman implied that the hardware selling price could fall faster than the total economic value created by each robot. So based only on the average selling price, the bill of materials, the estimated hardware gross profit per, you know, robot declines from 14,000 in 2025 to less than 8,000 by 2035. So what that means is that robot manufacturers may eventually need to make more money through software maintenance, fleet management. It's probably reoccurring subscription. this is going to be a whole model, you know, there's going to be whole new models that will emerge from this because the hardware just essentially becomes the entry point the same way you look at Apple, right? Like Apple makes a lot of money from yes from their phones by selling hardware, but a lot of their revenue comes from services that they, you know, give out. So like if you look at Tesla as a car, what do they make a lot of their money in? They make a lot of their money in subscriptions like the FSD software. So this is why robot as a service model could become so important. Instead of asking customers to buy you know hundreds of thousands of expensive robots you know a robotist company could charge a monthly fee to cover the robot you know uh robot and at the same time they're also getting data from you know from these robots as well as it gets deployed. Now this may like have some negatives as well like they have to go fix it and they have to like maintain it and like all that good stuff but this is where like utilization will become one of the most important metrics in the industry because the strongest evidence currently in this in that the industry is progressing is that we're actually seeing real deployments commercial deployments like figure AI um figure O2 which is the model robots spent more than a year working at BMW's Spartan factory. Um, according to figure and BMW, the robots uh operated uh 10 uh 10-hour shift, 5 days per week. They loaded more than 90,000 sheets metal parts, accumulated more than 1,250 hours of routine, and supported help support the production of more than 30,000 BMW cars. Now this does not mean like figure you know build 30,000 bytes of it just means that it perform like one specific task handling like one specific job but that one specific job will eventually turn into two into three into four and so on and so on. >> Did they share numbers of how much more efficient that was than human labor? Did you say that? >> They I don't know off top of my head. I'm sure they're also they're also deploying more robots like their new models into the BMW. So like I wouldn't be surprised like it doesn't make sense for BMW to deploy these you know robots if it didn't actually bring them economic value you know cuz all these are configured to very specific like BMW look at the company as a whole they've been around for hundreds of years right or 100 years or something and everything they've done is like you know they have the manufacturing like down to the tea so like in order for them to you know add in a robot it complicates their entire workflow so I'm sure they're getting a a lot of economic value from all this. So now agility robotics is another example like the digit robots that I showed you. They're being used inside logistics and other you know manufacturing environments. Um they're what's cool about them is they has actually built a factory in Oregon designed to produce like 10,000 digit u robots annually and what's even cooler is that 75% approximately digit parts are sourced within the United States which means we're not dependent on China we're not dependent on Korea we're not dependent on Japan for all these supply chain issues and just to give you a quick example remember when we blocked off China from accessing our own chips. Now, now AI war, we have largely won it so far, but as of like the next war is going to be robotics and China could easily cut us off to get ahead and to dominate the market. So there that digit company that I said is actually merging through Churchill Capital Corp sometime in Q4 and the ticker will come out as AGLT. >> So >> yeah, >> and and it's and you can actually invest in this company, you know, through through some um actually through this uh ticker symbol that's already out. So um there's been a huge huge speculation with this stock specifically. Um so this is one way to invest in that. Um another company Tesla we all know like Tesla is following a similar strategy with Optimus. Tesla is installing it like first gen like Optimus production lines using like real world AI vision battery and manufacturing and all that. Um the bull case for Tesla in my opinion is not it can simply design a human or robot but it could train the AI design chip build a f build the batteries manufactured robot at scale deployed inside his own factories and improve the data using like collected through these deployments. So they can literally vertically integrate unlike any other companies and that is exactly their moat because if you look at Elon and what he's been able to accomplish data center just data center buildout in the last like year like in the last 4 months he is building data center colossus one was built at 120 days that Colossus 2 was built in 90 days. This is like it takes years for other companies to build these data centers. So he can vertically integrate and build everything in house and even he doesn't even need customers. He has like enough work inside Tesla to like deploy these and you know make them better and do all that. And so Amazon if you actually look at the chart so look at this chart. Amazon currently has um how many how many do they have? This this chart is compare comparing Amazon employees and robots from 2013 to 2025. Um this is like from last year. So in about few years robots you know employees will be overtaken by robots at Amazon. That's how much they are deployed. And 2025 that was last year they're deploying so many robots in in in you know this year alone in all their factories and warehouses. They're building out Amazon pro like their own manufacturing and all that. So, they are massively ahead of this and they also Yeah, go ahead, LG. >> You know, one thing we've talked about a lot on this show is that crypto is quickly becoming a huge part of the global payments infrastructure. And nowhere is that more obvious than in Asia. But if you're actually running a remittance company or a payment business, you know that the hard part isn't moving the stable coins. It's dealing with local banking partners, compliance, liquidity, and all of the operational headaches that come with sending money into places like India and Southeast Asia. That's why today's partner is Saber. They give payment companies stable coinpowered infrastructure to collect and make payouts across Asia without having to build all of that complexity themselves. They've already processed more than $3 billion in transactions across 40 different countries. So, this isn't just a concept. These guys are actually doing this for real. If you're building payment infrastructure or expanding into Asia, make sure you check out saber.Money. >> What's this? What What is I don't understand this number cuz this says number of thousands. So, this you're telling me they have 1.5 employees, but in 2025 they had a million robots. Yes. Well, one well 1.5 million and 1 million >> 1 million. So a million robots robotic units. So that means like like an arm in a factory is a robotic unit. Okay. Okay. Okay. That's I was like I was like this doesn't make any sense. You know like I a million so they have a million robotic like things that work. >> Yes. And I'm assuming in one typical Amazon like distribution center or whatever there's probably like 30,000 pieces like that, right? Yeah. Correct. Correct. So this is just to show you that you know a lot of these companies are deploying like not a human or a robot but robots just in general. Um and what's cool about Amazon is they introduced an AI foundation called Deep Fleet um to move like how robots move through his fulfillment centers and Amazon expects them system to improve travel efficiency by 10%. Now 10% may not sound dramatic, but this like it creates enormous savings for Amazon specifically because Golden estimates that wider robotics adoption could create $72 billion in cost for savings by Amazon by 2030 and it could add $240 basis points in operating margins to his upside scenario. So, these are like Amazon will be able to save a lot of money by deploying like these robots in inside their inside their factories. Now, now I'm not gonna I'm not going to go into other companies and whatnot. I think I want to talk about how can you make money from this. I think that's what we're all worried about, right? >> Yeah. >> Exactly. Um so the I think the safest way to gain exposure is through investing in companies supplying the components and computing infrastructure and software to you know multiple robot manufacturers. Um like a like if one robot company fails the supplier can still to other can still sell to others. We have no idea which humano robot company will eventually survive in the in the next 10 years. We saw this happen with electric vehicles, right? Investors once believed that EV cars like startups like the could become the next Tesla like look at companies like Fisker or Arrival or uh Canoe. There's the list goes on and on. They eventually went bankrupt or shut down. The same thing happened with during the com bubble, right? Like hundreds of companies disappeared. But the companies providing chips, networking equipment, cloud infrastructure, payments, they all survived and they continue to support the business. And we also saw it in software like in smartphones. Where is Nokiia? I mean, they still I own a little bit of Nokia to be honest, but um where is But that's a AI AI infrastructure play by the way. Uh but where is Blackberry? Where is like Palm? Where's ATC? eventually like Apple and Samsung overtook the market, right? Cuz because investors who try to predict the winner like early winner, it's so hard to do that because the market doesn't even know what's coming. So I think robotics will experience a similar shakeout. There may be dozens of humanoid companies but only few of them or you know handful number of them will eventually reach mass production and that is why like you want to focus on the supply chain. Melvin, I have a question on that because that's an interesting analogy back to like the cell phone makers and providers as well is that you know there were some some companies that were in market and they're no longer the companies you think about but they had products in market though, right? like BlackBerry, that was their market to lose. And they literally in the Blackberry movie, they like showed when they were like watching the Steve Jobs thing and they were like, "Fuck, like we're totally fucked." And they knew right away. But they had they had they had a the biggest head start you could have ever had for mobile phones and they were just never thought about removing the keyboard. It's the simple the simplest simplest simplest thing and creating a good OS that people could build on. They never did that. But um for this it's like we don't have products in market. You know what I mean? So it's it's like we're not even at the part where there is a Nokia or a Blackberry or a Motorola yet. Like we don't that hasn't even happened. Um but what I am curious on that question is inspires us is like who's do you know anybody who makes the robots for Amazon and for BMW? Like who's are there any manufacturers that currently do that? And this is an off you know off-hand question. I don't know if you you know that. I'm just curious if they have a million units. >> I don't know actually. I have no idea. That's a great fun. >> That's all right. That's a good It's okay. It's okay. That's right. We can we can research this in pro or something like that. But that's I'm curious about like where does Amazon how does Amazon build their factories, right? Is like that's how >> I think they probably have custom like um companies that they work with and that they design and like similar to like custom chips, right? Like I think it's probably something similar. >> Yeah. Yeah. Okay. Well, we're here to talk about humanoid robots which are which are estimated to have like a far greater impact anyways, right? And is also something that is going to have more of an iPhone chat GPD moment than just a freaking random arm in a in a factory. So, I'll let I'll get back to you uh on this one. Is this where Tesla could be the Blackberry? Like is Tesla gonna make him millions of robots and then somebody else comes in and makes something better? >> No. No. I think Tesla, Elon, as long as Elon's alive, I don't think anyone will beat him in in in what he does. So, I think Tesla will be like one of the big big big player. But, you know, like if you look at the chart that LG just put up, you know, you have like a lot of different moving pieces with that goes into robotics. So each layer like essentially creates an separate investment opportunity for you know those of you watching. So if you look at Nvidia is like the clearest bene beneficiaries because it participates across almost the entire intelligence stack. Robots can be trained on Nvidia GPUs. The physical environment like we talked about can be simulated using like Nvidia's Omniars, um Isaac Sim, the Newton physics engine, Nvidia's Groot models can help doubers create robot like skills. Um they have a bunch of products that I'm not going to get into, but they that means that Nvidia can essentially potentially make money during and during and when the u you know robot is deployed. Now physical AI therefore create three different uh computing markets around the same robots for example like Google right is also becoming believe it or not is also a important player through Gemini robotics Google's deep mind models can also reason about like physical environments and control uh humanoids um control humanoids um track multi-steps run locally on you know robotic hardware um these models model layer may eventually work similarity similar similarly to a smart mo smart smartphone operating system like the OS. Um some robot makers will build their own intelligence stack while others may use models and you know develop platforms created by Nvidia, Google or Tesla like I'm sure they have their own you know that that will you know that they will like rent out you know essentially like get as a service. Um and as I said before Tesla is basically attempting to vertically integ integrate both the robot and its intelligence. You have figure um these are p figure is a private company. Um agility right? Agility we talked about digit robots. You can actually you know invest in that through CCXI um as well. Now the winner you know may not be like one universal model like industrial robots, warehouse robots, healthcare robots, hustle robots, there's al all like different robots and they all need like you know major computing power and that's you know I want to go back to my Nvidia you know thesis. Nvidia basically has all that in house. So like Nvidia will be like one of the key key manufacturies of this cuz if you actually go like um if you actually go to the next slide um over um real quick, Goldman actually estimate that each robot could contain approximately 3,000 to to more than 6,000 semiconductor content. This includes like 1,500 to 4,000 high performing computing modules, 750 to 1,50 in analog and mixed signal chips, 600 to to 800 in edge memory. So if Goldman like is correct about their projections, which you know, like I said, their numbers are way lower than what it is. I think simple math implies that there's a 19 to39 billion just flowing into the semiconductor demand from the robots. Now this doesn't even include like the chips used to train the models or the networking and storage infrastructure. Like there's a whole whole another side to this. Now another important thing is memory. As I said before, memory is a key key player is a key theme in this in this era because if you think about it, the robot needs edge memory to process sensor data locally. The cloud infrastructure training as a model requires DRAM, HPM and storage, you know, which is like your NAND. the company operating the fleet, you know, like your digit or like an different company also need to retain and analyze like thousands and thousands and millions and millions of videos, you know, and that that all have to be stored somewhere. Um, Goldman actually argues that a robot could eventually generate vastly more daily data than Corpus used to train a frontier language model, which means this this is crazy. This is how much there's going to be bunch of money flowing into memory because a robot without memory is useless. The same way I say chips without memory is useless. So you can invest that through companies like Micron which is one of my favorite ones or Samsung or SK Highix. That's a good way to uh get exposure. Now I want to go back and you know talk about some of the companies that may not be that obvious, right? So and actuators are another major opportunity. So Schoffler Schoffler says an average humanoid has approximately about 25 to 30 joints and everyone every 1 million humanoid robot could require 25 to 30 million join system. So 1 million times that by like 25 to 30 joints. That's a load a lot of lot of join systems. And these all have to get produced somewhere. And if you take the math from Goldman, 6.5 million units, times that by, you know, X amount, that's 160 million potential joints that needs to get manufactured, which is which is insane. And each think about it, each joint requires motors, bearings gear power electronics control system. So imagine in a world like one like that's that's the dilemma right now. one could turn into like 30 40 different supply chain like demand in the supply chain. So if you look at companies like if you go back to the um slide previous slide you can invest that too. Companies like Harmonic Drive or Leader Drive and Regal um Rex Nord could all benefit. Schaefer these are all companies that are that are like um that are benefiting from all this. And but what's what's interesting is you know when you think about actuators like actuators needs to be powerful enough to lift weight precise enough for you know like lift heavy weights and also it needs to be like precise like delicate movements and whatnot. Um, it also needs to fit inside a humansized body. And there's a lot of what I'm trying to say is there's a lot of moving pieces in this. And those pieces cost an enormous amount of money. And one of my favorite I think one of the favorite companies that I like in here is like Shafler, which is um because it participates on both sides. Um, it it is testing like robots inside its factories while developing bearing gear boxes. motors, sensors, you name it, they're they're doing it all. Um, and u, you know, one caveat I want to give to our listeners is a lot of these companies have ran up quite a bit, but because market doesn't know when this is coming, we go through these phases where, oh, robots got cool, so let me like pump some money in it. Um, now they're in in a six month or in two to three months, there will be another wave. Then those stocks going to go higher. I know I'm not saying invest money a whole lot of money into robotics. I'm saying keep an eye out for it. Invest a tiny bit amount on the dip and on the runup so you're like kind of invested in because these these companies like that like make the joints and gearboxes actuators they're not going anywhere. There's no different companies that really makes these other than the leaders because it takes a long time to develop these companies to be like this. >> And Melvin, I want to wrap this up. Can you just circle back on kind of the economics of these humanoids because this is like this huge buildout and everything is also like that's always going to be the question I think and you kind of mentioned this earlier but like like what is just remind people like what the cost is of one of these robots going to be and then like what's how do these companies plan to break even you know because that's even that's a big question with AI right now is like well even Neoclouds right we've done that it's like well let's how many years does it take to actually make money with everything that they have to spend and for all these frontier models to actually now they're finally showing revenue so That's positive, but how do you estimate something like that? And you know, we saw earlier, right? You showed us the TAM numbers, right? Which is great. It's great to know that this will be a $7 half trillion industry. Um, but you know, I guess how do you make money? Who who's making money and how do they make it with all this investment, all these components? This is, you know, it almost feels way more complicated than just making frontier models. >> Yeah, it's it's actually very complicated. And I think let's go back to the point that I made about the robotic cost. I think cost is extremely high right now because average cost right now is around 42,000. Um and that'll you know eventually come down to 21,000 by 2035 which means that you know the margins of these companies you know it will will also improve because right now these parts are hard to source and all that. I think the value in this in in this play is going to be all about the supply chains. That's where you want to be focusing your money and your attention at because that that's the same reason why like if you look at the AI trade like bottleneck trade what what does that even mean? Bottleneck trade means that the supply chain there's a bottleneck in the supply and whoever can fix that b you know bottleneck will you know their stock or like whoever has that component will greatly benefit from this. So these companies that manufacture these companies like this, whoever can you know build the fastest, get the margins, supply to these big manufacturing companies like Tesla and all these all these robot makers will be the benefactory for all this. So it's but it is hard to figure out who exactly wins. That's why you have robotic ETFs. Go if you're not sure of what to buy, go go buy ETFs. There's a bunch. There's a bunch. Are you going to put Robbo strategy in your Milkwood Pro portfolio? >> I might. I might. >> This is hard to resist after talking to Scott the other day and I'm going to be not going to not going to lie like he's very compelling and he's and guy even though he's out of research and I'd love to get Andrew as well on the show uh to to talk to him as well cuz uh I feel like this is a theme we're going to keep revisiting. Anything else that we need to know, Melvin, before we sign off? Um just that that image, you know, that LG put up. Just use that as a way to research. Um yeah, this would be Yeah, you can screenshot this. Go do some research on >> We got to make our own and put Milk Road on it, man. Don't screenshot this. Wait, Milro's going to put out our own image like this. Okay, just wait for that and use that and send it to all your friends with like a referral code or something like that. >> Yeah, exactly. We will definitely >> robotics cheat sheet. That's what we need. No, but this is wonderful. And I said this last time too that I think like even just looking at this list is really helpful to understand. I wish I wished for AI that 3 four years ago somebody showed me this for data centers, right? Like this is what I wish I had going into the last couple years is just like a list of here's everything you need. Here's who's making chips. Here's who's making memory. Here's who's making uh freaking cooling. You know what I mean? Like that kind of stuff. Like I needed this kind of guide. Um and it's really helpful for for robotics as well clearly. And just one last thing, I already own a robotics company in my Milkro portfolio. So like, you know, I'm making some plays and I want to make more plays as it arises. That's why I'm making this episode just to just to put this on your radar, guys. >> Mhm. Mhm. That's really helpful. Well, Melvin, another great one. Thank you, sir. Thanks for listening to Milk Road. If you enjoyed the show, make sure you like and subscribe. And if you're struggling to find winners in the market, that's exactly what Milkro Pro is built for. Our analysts have called some of the biggest winners early and Pro lets you see what they're buying next, every trade they make, and the research behind every position. Check out Milk Road Pro at the link below. Everything you hear on Milkroad is forformational purposes only. These are our personal opinions, not financial advice, and we may own some of the investments we talk about. Always do your own research and make the decisions that are right for you. See you next time.