We’re About to Need Way More AI Compute Than Anyone Realizes

On one side, everything coming out of AI continued to tell us like uh the demand got stronger, but on the macro side is where we were starting to see problems. What's up everybody? It's LG Ducet here and welcome to Milk Road Stocks, a daily market show that still can't figure out if slow September means a cool down of the market or just that I've been held back in grade 6 yet again. Today is September 8th, 2026, recording on September 7th. Which way, Western market? The macro looks pretty uncertain right now, but the fundamentals are strong in AI, showing a proof of concept people have been desperately waiting for. So, in this week's episode, our three top stock analysts give their vibes on the market before diving into the AI agent explosion that's getting hotter every day, and then they spend a big chunk of the episode talking about robo taxis as we have both Uber and Tesla bulls in the arena to talk about who comes out on top. This week's episode of the rollup is a little different. It's a day earlier than usual and instead of Vincent, we have Martin who invests both in AI, crypto, and even software sometimes at Milk Road. I also had to skip this one uh because it was Labor Day and had other things to do like hang out with my family and do some work in the yard. But it's very interesting. Make sure you listen because at the start, Martin talks about how he sold 10% of his portfolio already because of the macro uncertainty. If you want to hear their actual moves, we've started a new series just for pro members called After Hours, where after these shows, uh, these guys spend an extra 10 minutes discussing their portfolios, showing the screen, talking about their moves that they're actually going to make. So, you can check that out for just a dollar at the link below in Milkroad Pro. And a reminder that our podcast today is free and that it wouldn't be possible without our partners at Saber. Money, the stable coin payments platform built for Asia. Keep an eye out for more information about them later in the show. Gentlemen, we got a new lineup today. No LG, no Vincent. And we brought in Martin who's actually going to be here, I think, every week moving forward. Uh, and uh, so this is this is the crew plus plus Vincent in coming weeks. But, uh, Labor Day, we're recording on Labor Day, so markets are not open, but uh, let's hear your guys thoughts on the markets anyway. Uh, Martin, let's let's start with you. Um, what do you think's going on in the world of markets right now? >> You mean AI or like markets generally? Just generally, how are you feeling about the markets? What are you thinking? >> I sold some positions because I'm starting to feel a little worried and my cash positions were just so so small that I I didn't feel comfortable with all the macro uncertainty, you know, odds of raising heights uh spiking again. And like just in general, the the the markets are at alltime highs and I don't think that's the moment when you want to be all in. So I kind of sold 10% of my portfolio. I want to have more cash when the correction or whatever it's going to happen next. But like in general, I think right now we are going through a just a small correction in AI trade. But I don't see a reason why this this whole AI trade should stop and like you know there are still pretty interesting opportunities I think and the current valuations are like uh the I trade is you know gone but I don't think that's the case. Obviously there are some players that they might not do uh that well but I think we are on the lookout for the ones that they do. So, um, yeah, it remain bullish overall. >> Main's bullish, but sold 10% of his portfolio. All right, >> yeah, it's definitely bullish sign. Um, anyways, um, happy Labor Day everyone. Um, I thought, you know, we were supposed to get a day off, but here we are >> recording a uh, but so obviously the casino is closed today and the week just got started. So let's you know talk about like what happened last week. Um last week I feel like we got a lot of uh interesting news. On one side everything coming out of AI continue to tell us like uh the demand got stronger. Uh we had Dell earnings um absolutely ridiculous AI server growth. Broadcom had their earnings too which told us that um custom AI chip revenue doubled could double and then double again. Um Snowflake had their earnings too. I think they've jumped up 23% basically telling us that um there's you know acceleration in software usage um across their um company. Um then on top of that we had a bunch of model releases. Um we starting with Google um released their Gemini 3.8 flash model. Meta releasing their Muse uh Spark 1.3 I believe and finally then we had Open AAI releasing their um GBT6 Astra. Um and then on top of that, you know, we knew that, you know, Nvidia was going to acquire Hugging Face and they um acquired them last week as well. Um so we had a massive massive, you know, news coming out of the AI side. But on the macro side is where we were starting to see problems. Uh oil is back near like 95 93 uh because of Iran conflict. Treasury yields are once again rising. labor market um market came in significantly stronger than expected and because of all that tension and everything that's happening in in the macro side the odds the rate hike odds have basically jumped to I think it's around 60% as of today now do I think they're going to raise rates no because I I think we still have a lot of data sets coming between now and September 15th is when they're 15th to 16th is when they're having the meeting um we're going to have like PPI on September 10th um we're going to have CPI on September 11th with which I think like that should al alleviate a lot of the worries in the market and Fed's just going to end up holding rates. So, as far as that, I think we're going to if if they do hold rates, I think we're going higher. If somehow they do hike, you know, it's going to be a problem for the AI um bull run. So, that's that's where how I feel about the market right now. >> You know, we're about 50/50 odds. Uh so, the market basically has no clue. Um I I was listening to some stuff over the weekend as well and I don't even know that a 25% yeah rate hike is going to matter to the AI trade because it's growing so fast that like 25 bips really doesn't make a huge deal when these things are growing at you know multiples uh you know every year. Um, so it actually may not matter that much if they raise once. Um, because I the only issue would be if they went into a like a rate hiking cycle. Um, but I don't I really don't think that's going to happen. But, uh, I I I side with you that I do not think we're going to end up raising rates, but even if they did, I don't think it's going to matter that much. All right. What are my thoughts on the market right now? You know, Martin, you you when you sold last week, uh, or took some profits, um, I was like, you know, does the markets look that bad? And I looked at the NASDAQ and I said, "Okay, it's already gone through a correction and it has not gone back to all-time highs." Uh, S&P has gone to all-time highs, but it's kind of just sitting in chop pattern. I don't know. To me, I need to see a runup like this before I'm starting to think that we're going to have another big correction and like the one we had in the NASDAQ here. And we just have not had that. I know we've kind of come back uh after the correction that we did have, the unwind that happened in July. Um, S&P went a little bit higher, but not much. So, um, other than the macro side, I don't think we've like run too hot or anything on on assets. And then to me, it feels like we've probably already hit the top in oil. Um, and, uh, and at least the expectations of the rate hike have come down since last week. So, I don't know. I I feel a little less concerned on the markets. I think we've got a lot of things that are telling us that there's just so much need for the AI infrastructure buildout still. Um, if you look at Agentic use, uh, in the last couple weeks, in the last month, it is like I think more than I think it's two or 3xed in the last month, the amount of agent usage that's happening. So, like we're finally, we've hit that inflection point. We've been talking about agents for a long time in the show. In the last month, I think you're really really seeing this thing go through the roof. Grockbot obviously a big reason for that. And then just, you know, additional models and and kind of UX getting a little bit easier to do this. We're starting to see that become the new Martin, I think it was you who called this was just like labor as a service instead of software as a service. Um, and I think we're just kind of starting on this and I think that's going to drive so much more demand for compute because basically what's happening is the more that companies use agents instead of them like hiring humans, they're still going to hire humans, but now they have a budget that goes towards uh basically labor as a service. So tokens instead and that's expanding in basically every single company. And so, um, we're starting to see that, um, the demand for token usage, the demand for compute is just, I think it's going to be way, way, way more than most people realize. Um, and so I just think we need to build so much more compute, it's not even funny. Um, and this isn't even taking into account the amount we're going to need for physical AI. Um, and let's start there, actually, because we've had some good news, uh, or some interesting news here. Martin is a big Uber guy. Uh which is why we wanted to bring him in because Melvin and I wanted to to make fun of him. No, I'm just kidding. But >> can I quickly comment Gr before? >> Yeah. Yeah, sure. Sure. Because I think that for this whole agentic economy, the the key inflection point was January this year when open started because it kind of bring that whole framework how you can use agents >> and now Gregbot created something very unique because it is super simple for average person to create an agent. Before that it was such a pain like some of us would bought you know Mac minis and try to run open claw and then Hermes and like man like that experience is just a pain and like and yet like there were probably just you know few hundreds people uh thousands people doing that and using agents but there is 7 billion people on the planet and now Grubbot makes it super easy for everyone to spin out an agent. And just you know whatever you do you can just create agent or you can just go to marketplace and just import the existing agent that someone is someone else is using and yet you can start using it. Super simple. I think this is the future that open AAI cloud and everyone else is just going to bring as well because that's what you need. That's how you like, you know, lever that next growth for your for your product that you will just be able to provide more usage for your like every everyday users pretty much. >> 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 Milk 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 positions they're opening, it's just a dollar in Milkroad Pro at the link below. I think OpenAI and Enthropic are both going to launch a similar UX to Grockbot in the next 30 days. Like, it wasn't hard to do what Grockbot did, to be honest. Like, they can build that UX pretty easily. uh and that UI pretty easily. That's not the hard part. Um and so I got a feeling they're going to launch that pretty soon and you're going to be able to get that same setup anywhere. >> It's actually pretty painful to build uh these days because one day you are going to build something in cloud, the next day you are going to use open AI because now they have better model and then the next day you are going to pivot and use Grugbot. So it's like oh man what is going on? So I think the best version is that you just need to think about uh a way how to build your setup that you are kind of model agnostic. So you can just quickly you know uh just move and change your models that you're using on the back end and because now I think that bot really nailed that user experience and Slack and like everyone else is now probably going to do the very the very same thing. >> Yeah. Just to give you some context on the numbers that Kyle just mentioned about a agents, um, agents are using 14x more tokens today than in February. So that means they were at 0.51 trillion usage of tokens. Now it's at 7.3 trillion. Now humans has basically doubled as well. We were using, you know, 0.5 trillion and we're roughly using 1.4 trillion in the in the last like five months or since February. So, Agentic World is here and I think Grock is just going to continue to accelerate it. We actually saw proof of that. I think it was either Thursday or Friday when all the I think Grock got shut down because there were too many people using it or that that they ran out or something. >> No, they had an issue with their um not Colossus, but whatever their data center they have is in Memphis. Um and that that kind of shut down Grock plus some of their partners. So, I think Google and uh Enthropic as well. >> Yeah, there was a whole shutdown that that happened that day, right? So like we can see what happens if you don't build out the infra if you don't build out more infrastructure. This is what's going to happen and that's going to be you know it's not a it's not a pretty world that anyone wants to live in cuz I I I definitely don't want to not want to log in and not be able to use my agents. >> Well and one thing that's interesting is when we think about valuing companies um you you think about well there's an 8 billion there's 8 billion not 7 billion Martin but 8 billion people in the world. Okay. And so like what's the TAM of this product? How many people would want to buy this? And now you have to think about it very differently because there's eight billion people but eight every person not every person but eventually most people will have 10 20 30 40 50 uh agents each right and so now you're looking at a world where okay there was 8 billion people and now there's however many billion uh you know hundreds of billions of agents u and so like the TAM sort of changes I understand it doesn't change the amount of money that's in the world but it does change what they do and how they uh interact and what what products they use um and the velocity of spending etc. So it the world is changing actually in a very very fast rate. Uh but then you also have to think eventually you're going to have robots. Now you have 8 billion robots, hundreds of billions of of agents, trillions of agents and then I don't know how many billions of robots we get eventually. Uh and so the world will change even more whenever that happens. But that's probably a 2030 plus story. But again, let's go back to robots. Um Cybercap launched on Thursday uh in Austin. Um, so they've had the robo taxi network, but they launched the the vehicles that were built specifically for robo taxis, which is called CyberCab. Um, there's a bunch of videos on it on Twitter if you want if you haven't seen what they look like. This kind of the gold vehicle that they've launched. Uh, two-seater, no steering wheels, no pedals. Built specifically for autonomous driving. You can like put a a remote control like for a game device into the the screen that they have and you can just sit there and sleep or you can play video games. Uh, and it just drives you across town, which is pretty cool. It is currently, um, I think about four times cheaper than Uber and about half the price of Whimo in Austin at the moment. Um, so it is much cheaper. Uh, the cyber cabs themselves are actually way cheaper than even just like the Model Y. Uh, which is quite interesting. Um, by the way, to tie this back to compute also, now all these cars, they need a ton of compute in memory as well. So if you're modeling, hey, I hold a NeoCloud or I hold, you know, a compute provider, I hold a memory uh stock. The adoption of robo taxis is also extremely bullish for compute and and memory as well. Um, but you know, this is obviously a big deal. It's it's only in Austin, so it's still very small. Uh, but Martin seems to think that Uber is going to win the autonomous vehicle world. Martin, what's your uh what's your give us your bullcase, and then Melvin and I are going to rip it apart. >> Okay. Um, let's say it's the end 2027. How how many rides do you think or how many robo taxis do you think that are going to be life in US? >> End of 2027. >> Yeah, end of next year. >> Robo taxis. So, cyber cabs plus Model Y's plus everything else. Like I'd say by 2027, they're well above 10,000. >> Okay. Melbourne, what about you? >> Probably somewhere around there. 8,000 8 to 9. Let's just say it's 5,000. Doesn't matter. >> Okay. So um average average um Uber driver does about 20 25 rides per day. So let's say they will have 200,000 rights per day if your numbers are correct. And Uber does four 45 million per day. So we are going to be you know um like 0.5% in 18 months. Do you think it's going to eat like Uber's business? Like my take is that AVs are just going to expand the market because they are going to drive the cost down. So people who might be, you know, using buses or I don't know um they because taxis are too expensive, they they find some other ways how to commute. I think that's going to increase that market significantly and Uber is going to benefit from that obviously because they have the user base. They have 28 uh million active monthly users. That's huge. And I also think that Tesla, okay, there are two stories with Tesla. Um one is that they have cost advantage right now, but it's it's mainly due to uh tariffs because there are some Chinese operators that are able to produce the same cars at the same cost, like similar cost. at 30,000 they are able to also produce like robo taxis but they cannot be imported into US because they are banned so they have Tesla has that kind of production but like you know every like if you look at outside of the US like there might always be some tariffs or some importing taxes and stuff like that but robo taxis are becoming commodities it's it's pretty clear to everyone this point so that cost advantage disappears Um but if we just look at if we stay in US there are 52 states in the US right and you need to go state by state. So it's not like cyber caps can now be used for commercial use in Texas and in Florida but Tesla didn't even apply for um for uh this commercial license in California. And that license to it takes couple years to get that license. you need to collect miles, you need to do a testing and it's a long process. Or then you have states like Washington DC where they put into law that they are not going to just enable everyone to to use robo taxis. They are actually putting it into law that they will be just a certain amount of robo taxis because they want to protect the job market, the labor market. So there's a lot of people who would be put out of the jobs right away and they want to make sure that it's going to be a slow roll out and those people are able to find and accommodate to the new markets. So I think like all the people that think next I don't know in 3 years like Tesla is going to eat like the whole Uber's market share I don't think that's the case at all. I think the rollout will be much much slower than everyone expects because of the regulations because of the deployment at scale like I think today if there is 10,000 cyber caps from Tesla they are not able to find market for them right now because it's not just everyone can produce the car today but who is going to manage it who's going to maintain it where is it going to park how is it going to be you know charged and stuff like that that's not that easy and it's not just like, okay, you have a gigafactory in Texas that's going to produce >> I don't know what's the number uh I think something around >> every six seconds >> one car >> yeah one one cyber cap every I think it's 10 seconds something like that >> and if they cannot actually do it because if they do they still do not have the markets where they would deploy them today so I think the rollout will be much slower than everyone expects and in the meantime Uber is going to you know eat the world they just launched in London like the week before they launched in Tokyo. And so like >> Uber just launched in London. Is that serious? >> What are they doing over there? >> AVs. Okay. I was like, "Holy shit." >> Yeah. So So Martin's thesis is that Uber will become a marketplace for AVs. Uh and AVs will be commoditized, bring the prices down, which is great for Uber's business because it expands the TAM, the amount of people that will want uh to to use Uber because it'll be way cheaper. Um which I do agree with. Um I >> do you know who is also pushing a lot into AVs and who is building a model for AVs? >> Nvidia. >> Nvidia. >> Yes. So I was talking to um a guy that was at event uh and um he was talking about the so they put out a form on the day of the launch where you can like say hey I want to buy some cyber cabs. Um, and so basically what they're what they he thinks their plan is is they won't own the entire like cleaning and the infrastructure around these cyber cabs, cleaning, charging, even owning them itself. So they'll they'll sell the cyber cabs instead. They'll have their own fleet for for like high like super busy places where there wasn't enough demand to buy them, but they will start to like basically offload the cost of all this onto others. And I was like, okay, that's interesting. I could see some companies doing that. Like if you own like I don't know a big parking lot downtown, you know, Austin or something like you might you've got the space, you can get the chargers, you can have a little cleaning setup there, what a good way to add additional revenues onto your um parking lot business. But I was like, how many people are really going to do this? And he goes, imagine you own a store in like a little plaza. Okay, let's say you're a high-end like jewelry store or you're a I don't know, like a clothing store or whatever. Um you can own a few of these and you've got parking spots in front of you. um because there's a parking lot in front of your store. So, you can buy a few of these and start offering a service to your customers where you'll come and pick them up for free and bring them to your store, right? And so, like they'll go out and make money for your company during the day, but if someone wants to come to your store, you can offer them a service. You'll go pick them up. Same with like health care companies, like pharmaceutical companies, etc., where you'll pick them up and bring them to your doctor's appointment, right? And then you've already got a parking lot, so you can charge it. you can have a little business there where you can clean them and and have the charging because you already have power at your building. Um, and so you might own like three of them or five of them or something. I was like, that's kind of interesting. So, we just kind of talked about how like another way that they can scale that isn't just Tesla printing these and have to pay however much it costs them to to make these. Um, and in a way that they could like kind of scale it out throughout other businesses. And it's kind of just interesting to think about because if if the costs of these things become so cheap, um then the way that we use them is going to be very different than the way that we use Uber today. Um and so I just thought that was an interesting insight. Um I do agree though Uber is set up really well here. There will be different forms of robo taxis. Um, and some will be like they just sit in a little jurisdiction cuz like the one thing you got to think of, Martin, is not all of them are going to have as good data sets and as good FSD as Tesla. And I think that's where they're not going to be commoditized. And so I I do think there's going to be more companies that will have robo taxis. Um, I think many of them are very very far out from being able to work anywhere in the world. But what I do think will happen is you'll have like is it Zuks or whatever? The Amazon one. It's that weird looking car. It's not a very good model though. And so like maybe they'll have these very small geoence locations in very busy areas like super like right in the center of New York or right in the center of like I don't know Chicago or something and that's all they can do. But they'll just have a few of them in there. They're going to be super cheap. They kind of they fit that one thing and that's kind of it. But they don't leave that area. Whereas Tesla Cyber Cab because it's smarter and better might work across city and it can go on highways whereas some won't be able to go on highways because it won't be as smart. That's where I think it'll play out for many years. eventually maybe, you know, we get to the point where you can have a few of them. But does Tesla kind of own it by then or does Uber own it by then? I don't know. We'll see. It's going to be interesting. But I do think either way, it doesn't hurt to own both Tesla and Uber because they're the clear leaders of it all. Um, you got Whimo, who's kind of like doing their own thing. They were part of Uber, which was bullish for Uber, but now they're leaving Uber. So, we'll have to see who else can scale and who can actually create FSD at the level that Tesla is. Right now, I don't think anyone is even remotely close. So, I I do think Tesla has a multi multi-year head start uh in front of everyone else. Um except for the fact that what Uber can do here, and this is this goes on your bullcase, is again if you have these like lidar type um AVs that are trained in a certain small area and they can get a bunch of these then Uber can kind of be the marketplace for that and then it can fit a full jurisdiction. Um but or geoloccation, sorry. But yeah, I don't know. We'll see. I think I think owning both is a smart idea here though. >> I used to be an Uber bull uh to be honest. In fact, I owned Uber uh for for some time. Then I realized, you know, like but Martin's right. Martin's right to the point where Tesla is not going to roll out uh these taxis as fast as you think. You know, Elon's timelines are always off 2020. I'm not even but this is not I'm not thinking about the next three years. I'm thinking about the next 10 years, you know, or the next 15 years. You know, that's how market likes to think. um how is Uber's moat going to get um you know disrupted and I think Uber's biggest moat right now as you know um Martin said is their distribution they have a huge huge distribution channel 28 million subscribers or whatnot but Tesla can build the same I bet you because if if if you look at it Elon does no advertising you know but you know if you look at Axe you know that's his biggest you know advertising forum but indirectly he has never done advertising for Teslas or any of that matters and I think he can easily convince 100 million people, 20, 50 million people to download an app that takes 5 seconds. Why would you not download the app if it cost you significantly less? Like imagine imagine you go like I'm sure we all all have taken Ubers from the airport now how expensive that is as soon as you get out like from the airport. I would much rather sit there, download an app if it's like 20% 30% cheaper and it's gonna get me home safer because Tesla's has never had an accident before. And it's it's much much safer than every everything. And Whimo deciding to leave Uber, you know, is like the biggest thing cuz like once you get big enough, why would you want to stay in somebody else's um distribution channel when they can build their build out their own distribution and just have their own app? The interesting thing Uber has is the um is not necessarily ride share, but delivery, food delivery, right? I think it's in terms of revenue, it's the same size as their ride share and partnering with drone with the what was the drone company that they partner with. I do think that's a thing that Tesla's probably not going to go after. Um and so like there's a world where Uber I mean Uber definitely owns that. So that's interesting to see too. So it's not like Uber is a oneshot pony. like they do have other things and I think ultimately the TAM of just transportation in general is going to get so big because we can just you know like you said we don't need buses anymore if you can just get a car. I was like people are going to just rent these cyber cabs out just to like get away from the house you know what I mean like they're just cuz it's quiet they have like a TV it's comfortable like they're just going to it's going to be a thing you do you just go and get in these things. Uh but I think um people will probably stop cooking food and going grocery shopping. He said they're just always going to order food from restaurants because they can get it for like a dollar to get it delivered from a drone, right? So, I just think the world is going to change a lot from these from these um autonomous vehicles. It's going to be interesting to see how that plays out. >> I've always wondered why um Elon has never went after the drone business. >> Me, too. >> He's got everything else. Why wouldn't he do that? >> Maybe he's cooking something up. Who knows? >> Do you have one more thing you want to say about Uber, Martin? >> Yeah, a couple things like about the drones. I think uh what makes me excited is that I'm I'm going to probably order less food from the restaurants around because I will buy Optimus and he will cook for me. So uh that's actually against >> Yeah. But I think um you guys cannot just assume that Uber is not going to has access to any AVs. They're going to be a bunch of them. They already are. So like it's not like Uber is just you know humans like drivers versus like I don't know Tesla. It's not the case like Uber is they are even like promoting themselves as a hybrid solution. They're going to have they're going to have humans alongside the robot taxis and that's the winning strategy because if you look throughout the day on the usage of how many people demand you know right chairs right then probably like if it's um I don't know let's say Sunday uh 1:00 a.m. in the morning or Monday when there is no one out there, people are sleeping whatever then like the the request for the rights is pretty low and then you have something like Friday night when everyone wants wants to go out. So the demand is just through the roof and like okay how many cyber caps should I deploy in that market and you are either going to underutilize your fleet or you are going to just have wasted your cars. Yeah, but this is the part that you miss, Martin. So Tesla has two of the most sold cars in the entire world, Model Y and Model uh Model S. There's 9 million of them sitting around the world across every major city right now. Okay? And so at any point they can turn that on as a robo taxi as well. And so if the demand gets really high, they can send me a notification and say, "Hey Kyle, do you want to add your Model Y to the fleet?" And I can just say, "Yeah, absolutely." and then it leaves my garage and it goes and picks people up from whatever concert that they were just at. So Tesla is the only other company that has that. The other thing that's really important I think that that I think a lot of people miss is to get to the point where Melvin was talking about how safe the robo taxis are. Um to get to that point you need, you know, Tesla has 14 billion uh miles driven. No other car is even remotely close to that. Why? Because they don't have the scale of cars out on the road. and they can't get it because the the thing you just said, Martin, which is you can't have autonomous vehicles go out on the road in scale because it doesn't let them. But what Tesla has that no other car company has is they have 9 million other cars with all the same equipment collecting all the data driving around the road, but humans are driving them or they have FSD. It's not actually a robo taxi. Whimo doesn't have that. They don't have LAR on 9 million vehicles. No other car has that except for maybe like a Chinese car, but again, they're not being allowed into the US. Um, and so they don't have that yet, but none of the other competitors that are on Uber right now have the ability to get that amount of data to train their vehicles to make them that good to get widescale approval. Only Tesla has that. So that's the edge where like Tesla is far ahead of every other company and no one else can catch up because they can't scale their cars. Um, so they're never going to be able to catch Well, I'm sure at some point they can, but it's many, many, many, many years out. 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. >> All right, we've done a lot of talk on uh on on ads. Let's move on to something else. You want to talk about the models or is there something else you guys want to talk about first? >> Let's do models. >> Okay, Melvin, lead us off. >> All right, so last week we had um a lot of model releases. Um, like I said, we had Gemini, we had Meta, and we had OpenAI all releasing their models. Um, Google, uh, let's start with that. Google released their 3.8 Gemini model, flash model. What's interesting is they kept pricing at 0.75 million, uh, per input tokens and 3.75 per uh, million output tokens. The important thing here is that, you know, the models got better, but uh, without Google making it any more expensive. um flash models, you know, that's what they release are meant to be like fast and cheap and um you know, enough to use at a very large scale. So you can put them inside like customer service and search video analysis and internal software, you know, yada yada yada. Um and Google also in um improved their agentic video capabilities inside these models. Now uh models can understand and longer videos and you know process them efficiently. Um and uh that makes things like security footage and sport analysis and factory monitoring you know all these all these utilities can come out of that and now it's cheaper because of that and a lot of people assume like as soon as they hear cheaper models right they they think that um that means less compute demand that that is never the case but like if you make AI dramatically cheaper people people are just going to use this more often that's Jevans paradox at its core and that's what Google basically showed us uh with this. >> And I think what's coming with these flash models is if you want to use agents in like a way where it just fits within your like daily life, you need it to be fast, right? Like if I was to ask Claude to just like I don't know add something to my calendar or like tell me about my, you know, some email or whatever, like just a quick question, it it takes too long, right? Like it's got to load and kind of like takes, I don't know, 30 seconds to spit you out your answer. Whereas what Flash is doing is making it like it's a second. And so now what's going to happen, I think, is they're going to use these flash models. They're not as smart as Astra, which just came out from OpenAI, but they're pretty darn close, but they're way faster and cheaper. And I think that's going to be the Agentic uh product that they're going to launch through Android. Uh and I think Apple's going to end up doing the same thing. So by the end of this year, I think inside of our phones, you're going to have the ability to just like talk to your phone. It's just immediate response, right? Immediate response, immediate task completion, whatever it is that you're going to do. And so like picture Grockbot, but instead of like this UI where I need to talk to it, I'm just going to have agents that I can just talk to through my phone and it can immediately do things and then like respond back to me and I don't have to wait. I think that's going to be a huge huge huge product. Google and Apple are set up the best for that. I >> actually wonder what's their strategy because they let go not let go a lot of good talent actually left Google. So they were pretty much like okay we are giving up on competing with Frontier Labs. So, and if they are just trying to, you know, build a decent model that might going to be uh working for, I don't know, Gmail or Google calendar and that's it. Like, I don't know. Um, but I'm actually surprised that they are still sort of in the race still. >> Well, they've got Gemini 4 coming, which may or may not be something that competes. We'll see. I wouldn't date Google. You think it's going to be a flop? >> Yeah, I think it's going to be a flop, unfortunately. And by that time they come out um I think um um Claude is already releasing their model sometime this month. It's going to dramatically like you know or leaprog them. So I'm not I'm not really this is why I made my bearish case on Google a couple months ago cuz I don't think they they're even in the race. You know Chinese models are beating Google >> but they don't need to be I guess in this race. >> Yeah. They have the distribution right. They've got the phones they've got all the products. they've got all the users and so they just need this is why I think they're going the flash side instead which is there there's going to be different types of models right there's like your your genius ones and that's your you know your frontiers right now and that'll be used for things but then you need really fast ones that are still smart enough and that can be used inside of all the things you said Melvin like customer support and blah blah blah um so I think that's what Google's going towards >> yeah and then um on the other hand Meta Muse um released their Spark 3.1 model um and the biggest biggest thing with this model is like longer agentic workflows. I mean, I'm not going to focus on this model too much or we're not going to talk about it because it's not as sexy as like the Astra that, you know, GPT just just released, which is probably the biggest like model development um you know we had in a while. Um just to give you some context, Astra scored um 97.6% on Frontier Math uh tier 4, 72.6 on OS World, 95.9 on Bench CAD um and 100% on Exploit Bench. Basically what all that means is models is just getting better at advanced math computer use and engineering and cyber security tasks. But honestly like I think benchmarks are less important um than actually what um Astra is designed to do. Um the you know way open AI frame this is it's built to built to operate browsers spreadsheets CRM like systems terminals coding environment yada yada yada. So, the main takeaway from this model release I got, if you want to know if a model is actually good, just go on X and see who is talking about it. Like, seriously, that no, I I'm being so serious. And there was I hardly hardly saw anyone um you know, talk badly about this model. Like, it's it's it's a good release. I mean, there's internal testing benchmarks are still coming out, but so far this is the best model in in the world right now as of now. Jensen Jensen from Nvidia called it AGI. >> Yeah. >> Yeah. That was Is it Do we do you guys agree with that or where did that come from? >> I don't know. Like >> is he just trying to market? >> I don't know. Um >> I think he's excited to tweet out, man. >> Tweet out on us. >> Excited to be on Twitter. Uh it it is a beast of a model though and I think OpenAI is having a huge huge huge comeback and whenever we get their next revenue numbers, I think they're going to be an absolute beast. Um they they've been cooking. Uh so interesting to see and they're selling now more than a billion dollars in ad revenue too. So they're they're starting to diversify their their revenues. Um so they're they're becoming a beast of a company. >> And what does this mean guys? More tokens will be used. That means we need more inference compute >> now. That means more money is flowing through the infrastructure layer. The infrade you know all this is just going to create in like a crazy demand across the entire AI stack. And that's why I think AI is taking a with the semiconductor trade is a little bit taking a little bit of a breather. But we saw last week on Friday had a massive massive like um runup. Uh not you know massive but like you know a decent runup I want to say. We saw Bloom Energy go up. Nebas go up. We saw memory trade is back. You know it went up. It crossed thousand thousand bucks I believe on Friday. >> I think SKH highix is up 7% last night. >> Yeah. And Samsung. Yeah, their co Cosby is up about 5% as of today. >> Yeah, I I don't know about you guys, but I think that this bull market is going to go a lot longer than most people think. And I and again, I actually don't think that if we raise rates, I don't think it's going to matter. Um because again, the just the growth in this is so fast um that it just it's it's raising by 25 bips isn't really going to do anything. Of course, if they raise like 200 bips, okay, that would matter. 25 bips I don't think will matter at all. um we're seeing inflation continue to come down again as long as as the war gets solved or whatever gets figured out over there then um I really don't see the macro side being much of an issue and I just think as we talked about earlier like we're just getting started on the agentic side these models are getting so good and it's just kind of non-stop I think token usage is just going to go continue to go through the roof way more than we all realize um I think the market is underestimating how much companies are using AI and how much benefit it's bringing to their businesses. Uh, and I think it's a big reason why earnings are going through the roof, too. Um, and so I just think we are in a um, I wrote a tweet about this over the weekend, but like >> everything bull market. >> Well, we are definitely in the everything bull market, but people keep comparing this back to, you know, the the 2020 2021 bull market and theuh 98.99 uh, you know, market that we were in. And the thing is is like in 2020 2021 it was a subsidized bull market, right? like we were giving stami checks to the consumers and then the consumers were using that to buy dumb to trade and do whatever. So that created earnings were really really good in 2021 but it was subsidized earnings right we were paying people to go and buy So that that's not what we have today. We we have no subsidies we have no sty checks um and so that's not what's happening in '989 which was kind of the the tail end of the bull market. remember that was like a almost a decade long bull market. Started with infrastructure in like '94, '95, '96. Then like '97, '98, 999 it went to like everything uh anything that was to do with with the internet. And but that bull market was based off metrics like visitors to a website, right? Um which isn't a a an earnings or revenue metric. It was just like, you know, how long did they stay on a website? How many times did they visit a website? And like if your numbers were going through the roof, then everyone's like, "Oh my god, this is super bullish." And so like these the the multiples that these companies had were just insane. So if that was the case here, like maybe it was like how many people are downloading OpenAI app or like how many tokens are being used. Sure, even that that's not really a metric, but the metric that matters is revenue. And if you look at the model layer, the earnings layer, and the infrastructure layer, all of them have earnings going through the roof. Their multiples are nowhere near what they were 98.99. So we don't have this like fake bull market or incentivized bull market like we had then. We have an earnings bull market and I actually think those earnings are going to continue for the next year at least. Um and so I think this bull market continues for much much much longer. Do you guys agree? >> We keep seeing revenue growth that's accelerating. That's awesome. That's what you want to see with every startup. Well, should we call them startup or so or or like you know it's one trillion they are supposed to do IPO at 1 trillion so I'm not really sure but those are still revenues not profits right so it's it's a big difference because like they spend a lot of money I'm not even sure they might be profitable but just you know not that much as people >> Enthropic was in Q2 and the rumors are open I will be in Q3 but we'll see. >> Yeah. Yeah. But like what this tells me like the more we use it, the more agents use that and we see the acceleration of tokens usage, it makes me much more bearish on the labor market because I think that what this ultimately means is that either we I am going to be more productive so I'm going to provide like more offerings and it means I'm going going to eat lunch of someone else who is not using AI or either I'm going to cut cut my cost and I'm going to improve my efficiency by you know letting people go and again it goes against labor market. So like I think that the more this usage goes up, it's gonna like I think it's leading indicator for for the jobs market. Actually, >> real quick, I was at a buddy's house on the weekend and he works for the fastest growing cyber security company in the world. Uh private private one um and they have a this company um not only so basically they work with all the biggest companies in the world to help them figure out how to use AI in like a way where they're not going to get hacked, their data is not going to get shared across people within their company slash out of the company. Um, and their company also uses agents um, uh, a ton. They're actually a case study for Anthropic to go to other enterprises and sell what they've done because they built the fastest like agentic buildout um, that got I think like 82% use across their entire company within like it was less than 2 months. Um, so what he was talking about is the thing that they that's become really popular is it's called humans in the loop and it's basically it's using agents and it's using AI but with a human in the loop. this is like the the best thing that companies can do. And so what's happening at their company, which is seeing some of the fastest growth of using agents ever, is he's like they're still hiring like crazy. There's certain roles they're not hiring, but there's many roles that they are hiring like crazy now because they're becoming more and more profitable in what they do. And they want more humans in the loop on the things that they're building. And so he's actually like, "Look, our company is using AI and accelerator rate. Our profits are going through the roof, but we need more humans to keep those profits growing." Um, and so he kind of had the similar mindset to me, which is I I understand what you're saying. There is labor that's being disrupted here without a doubt. But I just think it's changing and now it's like we need humans in a different area, in a different type of thing. And so I don't think unemployment will go too crazy as a result because I think companies are just going to get more and more profitable, but they're still going to need humans and they're still going to be hiring humans as a result. Um, so I think we're switching what humans do. Melvin, you got some? Yeah, I think uh I used to have Martin's views up until like maybe three four months ago. I I thought this was going to get disrupted by a whole lot. I think this is partly caused by the leaders of the AI world like you have anthropic you know Daario you have open AI Sam Alman coming out saying you know jobs will be wiped out 60% of you know people will be unemployed white collar jobs is going to be gone but I don't think that's going to be the case at all. I think this is this is just going to create more demand in the labor labor world. I think people are just going to create more jobs. >> Right now we're in a productivity boom, right? And so if there's more people buying things and there's more people doing things and companies are getting uh growing their earnings, they're going to put that back into research and development and building new things and innovating and that means you need more humans to do that. So I think we are just changing what people are going to do. But I think because they're so earnings are so good and I don't think that stops anytime soon, they're going to keep hiring. Now, I do think the next time we go into a moment of earnings going down, assuming we hit that whenever, let's say inflation goes through the roof in the next, I don't know, 2 3 years or something and we need to go through like a rate hike cycle and all of a sudden companies need to tighten up. That's when we're going to see a lot more unemployment. But I don't think we're going to be in there anytime soon because we're having such a productivity boom that we just need to keep hiring because agents and AI are not fully taking over everything yet and we don't have robotics to take over everything yet either. So I think we're multiple years away from that. I do think though at some point whenever we get to the point where money gets um scarce again uh because rates go higher whatever um then I think you're going to see a bunch of companies let go a ton of humans and we're going to they're not going to come back into hiring. So then that's a problem. But I don't know. I think we're still a ways away from that happening. >> I don't I think we don't see the big layoffs yet. Well, like Uber just uh you know cut 10% of the workforce because so productive. >> No, I mean it's because if you are a company that's using AI a lot and you are probably at the frontier, you are gonna see a lot of opportunities. So you are hiring like crazy. But the company that you are replacing, you are eating their market share. They are not they are not like dying tomorrow and like letting people go tomorrow. You know, those are like survivors and they might going to be here like for a couple more years and just slowly dying. So, it's not going to hit the labor market like I don't know this quarter. But it it's happening under the surface. We just don't see it yet. But like I I still have like really strong conviction that this is going to happen. >> So are we all just going to be employed by Daario in the future? >> I mean he did say he said he's going to be the only private private company in the world left. >> Well he's going public in like a month so he's already going to be wrong on that. But >> I think a lot of companies are still in that learning and experimenttor like experimenting uh phase. Uh but I I think like probably maybe starting next year we could start to see that people are okay. We have built our AI factory and now we really can like let some people go and just focus on core our business. And >> now are you bullish on the on on those companies or you bearish on those companies? What's going to happen to the stock market if that happens? >> Yeah. Uh good question. I actually think that the like there are going to be you know all these hyperscalers they have the compute so they have a huge advantage I think they will just go much much higher and then there's going to be a gap between like all these like trillions companies and then there will be probably a much more like smaller companies that are going to focus on their niche only and it's not going to be you know trillion dollar market but it's going to be they're going to be good enough in their specific niche >> so I think by the time that that happens the earnings of those companies are going to be so good that the NASDAQ and the S&P are going to be so much higher, which is going to allow all the boomers that have not been able to leave the labor market finally leave. And then you're going to realize that, hey, actually the labor the participation rate in the labor market is just coming down because a lot of people shouldn't be working right now. They're already at the age where they should be retired, but they haven't been able to. And so hopefully what happens is the stock market just goes up so much that they're like, "Okay, sweet. Finally, I can take this time off." Also because you got to assume that if that happens that's a very deflationary world which means things are going to get a lot cheaper if that your your scenario plays out Martin which means a lot of people can work a lot less and so maybe unemployment doesn't actually end up going up because a lot of people just leave the labor market in in in general which is already slowly starting to happen. You're seeing the boomers who should have left the market years ago are finally starting to walk away and so um that's why unemployment is not going up like crazy even though there are people leaving the market. I I think since I live in me specifically since I live in the United States, I think I'll be fine. I don't know about you guys, but >> you guys gonna figure it out. >> Yeah, we're going to figure it out, man. >> I we already I already live in a social economy in Canada, so we've already figured it out. It sucks, but we we're there. >> You guys have a recession every every two months, but relax >> 100%. All right, we're going to wrap up here. Now, we got to talk about our portfolio. So, those that are listening, this is where we do the after hours show, uh which is for pro members only. So, if you're not already a pro member, go sign up. It's only $1. Uh, we're going to literally share screen right now and show Martin's portfolio, who he just sold 10%. We're going to ask him why the hell he just did that. Uh, and then we'll do the same for Melvin, although he didn't sell. I think he was buying. So, we'll talk about that as well. Uh, all right. Thanks everyone for listening in. We'll see you guys in the next rollup next week. And, uh, hopefully you join us over in the after hours. 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 Milk Road 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 Milkroad 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.