Why the AI Supercycle Will Last Until at Least 2030

If you've been in this trade, or even just watching the markets for the last 2 years, analysts keep saying we're like 6 to 9 to maybe 12 months away, and then it the whole thing kind of tips over, right? And so it's like we're going to guide up for 6 months, 9 months more in terms of hyperscalers CapEx, and then that's it. Then it's going to really slow down. And then the next quarter comes and it's, you know, they go, "Oh, it's 9 more months." And then the next quarter comes, it's 9 more months. And that's literally been happening for 2 years, which is exactly what Melvin was describing on the Melvin on the mem- memory trade. And so I think we're going to continue to see that. I don't see a slowing down. I don't know about you guys, but I really don't see a slowing down until like 2030, which like the thing is is you'd think you'd slow down then, but then that's when you have robotics taking off. And you're like, "Okay, well, now we need even more data centers and more memory and more inference and more, you know, compute." And so it's like, does it slow down in 2030? I don't know. The market keeps thinking this, but I just I'm starting to think it just keeps going and going. >> What would change that, Kyle? >> For one, I think the robots story is people are getting too excited about it right now. I think we're still years away from humanoids actually making any significant difference in the demand side of anything, whether it's like the parts, whether it's in compute or inference, any of that. Like, I think we're still multiple years away before we, you know, I think by 2029 we might have like maybe 100,000 human or robots, you know what I mean? Like, there's not going to be any real production going on until probably 2030s when it really starts to take off, is my guess, okay? So, you can't really include that in the numbers today. And so we're what? 4 years, 3 and 1/2 years from that. So, if the agentic economy didn't take off right now, cuz that's what we need, right? Like, I don't think we're going hockey stick curve on just like coding, and just like chatting to to ChatGPT. We need agentic to take off. And that's the thing where like it has not yet, except for on the coding side. Now, I guess the only other thing is like I think it's like 20 or 30% of companies are really spending almost all of the the spend on tokens right now. So, like there's also world of that diversifies out and goes to all the other companies, right? Um so, uh we just need usage to continue. And if that stops in some way or another, then that's the thing that says, "Okay, we're going to slow down." But I just I don't see a world where that happens at all. >> And so, I have I have many points actually on this. So, A regarding Tesla, just really quick to get a feeling for when physically AI, like robotaxis, humanoids really coming. AI 5, the chips of Tesla that Samsung is producing, just taped out, I think, a week or two ago. That was That's coming from a LinkedIn post from a Samsung engineer that was later deleted, right? I think he was not supposed to to post about that, but he did. Now, what taping out means is they now they need to tape out >> Taped out. >> tape out. It means they have to like test it for another couple of months. And until it goes into full production, there will be earliest beginning mid next year. And in my view, AI 5 chips are the core driver for robotaxis, right? Just that that gives you an and then humanoids are even further out. So, this gives you a feeling of how far away those things are from full >> robotaxis will work on the on the fours. >> They will They will >> Theorize now, though. >> Agreed. But in my view, but that's just a personal opinion, there's no no data point behind it. Elon Musk will use the AI 5 chips for the entire robotaxi fleet. I think he will use the AI 4 chips now to test like what what he's doing in the different cities, but the real scale fleet will be AI 5 chips in my view. Yeah, so that's that's that's the first point on that. Second point on when will the when will token demand potentially slow down and therefore the demand for all the infrastructure behind it? To me, the biggest risk, like kind of black swan in this, is when we reach ASI eventually, so super intelligence, and then the the super intelligent agents can figure out how to better design the chips, how to better design the software around the GPUs to decrease the biggest bottlenecks, to decrease the demand for memory inside the chips, for instance. Now, I agree that this is down the road and nothing for probably this, definitely not this year, maybe next year, not sure, right? But this is a a risk that is not priced in at all. That I see that could slow this down, right? To answer your question, Algy. >> 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 Micro Pro are all over it. They spent the last couple weeks making a lot of trades, getting out of some positions, and 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 are having in their portfolios, what positions they're opening, it's just a dollar in Micro Pro at the link below. >> Real world assets like funds, treasuries, and private credit are still running on rails built decades ago. Gated, paperwork-heavy, slow to settle. Everyone's talking about tokenizing them, but far fewer can actually do it, and do it without cutting regulatory corners. Securitize can. It's the SEC-regulated infrastructure bringing real world assets on chain. Nine years in, native tokenization, not wrapped, backed by BlackRock, Morgan Stanley, and Cathie Wood's Ark Invest, and chosen by the New York Stock Exchange, VanEck, BNY, and Apollo to do it at scale. It's the regulated bridge between traditional finance and crypto. Tokenize the world at milkroad.com/securitize. >> So, similar to the um I guess what we've discussed in recent weeks, as but I mean, it's a similar argument, it's not the same thing, that basically AI getting better and smarter would reduce the need for compute. And what you're saying, Vince, is a similar argument that over time AI would create a far more efficient way of manufacturing everything that it needs, which would then slow down the need for all the capex that's being spent, right? >> No, because the more efficient it gets, the more we're going to use it, the more use cases unlocks, the more compute we need. That's not going away. >> Yeah, to a certain degree. >> Yeah, I was going to >> I fully agree on that, right? It's the same argument that Kevin Bacon makes now around the open source models and the cheaper the token tokens get, the more it will be adopted. Right? I fully agree. The question is, can the redesign of of of the bottlenecks and and and all the imports like memory be so big that it actually offsets the higher demand? That's that's the question there, right? >> But also to get to like AGI and that to the point you're talking about, we need a lot more compute even to get to that point. You know what I mean? Like we're nowhere near there yet. So I think there's still a lot to go. >> That's why I'm saying it's a black swan, right? >> Yeah, yeah. We're also like and I think you brought this up a long time ago, Vincent, there's also the demand from governments and and you know, nations that need to build up their own AIs, which is happening and you talked about this, I think Melvin and I made fun of you cuz we said European countries will never do this. But I'm actually starting to become a believer that I think they'll at least try, which is what you said as well, because I think there's you know, so there's more demand to come from that, right? Cuz a lot of them are just starting. So >> We see news around that. Japan, South Korea, UAE, of course. But also Europe. Mistral, yes, it's a small lab, Melvin, I know that. It's not comparable comparable to the US peers, but still they're raising couple of billions and the French government is supporting them as well. And it's bullish for us investors, right? Because in the end it tells you more demand for AI infrastructure, if whether it will succeed or not. >> You know who's the biggest AI infrastructure builder in Europe? Any guesses? >> Nvidia. >> Nvidia. >> Nvidia. Yes, sir. >> [laughter] >> What Wales was the news now, right now? >> Yeah, the Wales and they just hired two more people in India, so I think there's a data center probably coming there. there soon, too. So. >> And yet it's down 8% today. >> Yes, sir. >> Kyle, one point on your on your argument saying that we do not see agentic um demand beyond uh coding. Yes, I agree, but if you look at Semi Analysis, the company itself, right? They went from a spending budget from a couple of hundred thousands, I think to an 11 million dollar run rate for this year. So, they're spending 11 million on APIs in the end, right? And and I think to me that is showing you where we're moving to. Once companies start to realize how to actually adopt AI in a in in in a way that's valuable, then there is no limits. There's no limits. >> So, I I agree. The problem is we're not at that point. Most companies don't know how to do it at a good ROI, and plus tokens are still too expensive. So, there just might be this window. This is my kind of black It's not a black swan, but this would be my answer to LG of like, what would stop this train? It's like, if that window doesn't close, right? Like if if if we don't figure out that out, and we don't get that ROI, companies going to be like, "Fuck, I got to pull back here, you know?" I don't think that's going to happen cuz I think we will figure it out. Um but there's just a chance that maybe that does happen, you know? That's That's It's what I talk about risk, that's what it would be. >> I mean, there are many like Coinbase, Robinhood, also Uber this week. They all are showing already in their in their earnings reports, if you look at the transcripts or the recent posts of the Uber, I think COO it was. They went like the first 6 months of this year, they went through a testing phase, right? They were just telling their people, "Hey, go spend, and we'll figure it out." Then we reached the top where they're like, "Okay, this is this is way too expensive for us." And now they're moving into the efficiency side of figuring out approaches of how that actually works. And I think I agree this will take a couple of months, maybe years, but this will go like this process companies will go through. And and this will just increase the demand for for agents moving forward. Right? It's the same story right? We we have the same story. >> Yeah, 100%. Keep in mind that majority of businesses in America are small mom-and-pop shops or like smaller companies with like 20 employees not which have not tapped into the AI at all, you know, so there's a huge huge demand that will come >> And for them it's much easier than a big corporation. >> Exactly. >> People keep trying [snorts] to compare this to the dot-com bust and the the thing they always miss and actually LG you had I forget who it was on the podcast a couple days ago and he talked about this where there was that gap in between like creating the infra and then actually using it because you had to go and like buy a computer which no one had at the time. Then you had to go and get internet set up which means someone had to come out to your house and install it and then like plug it into your phone so you couldn't use your phone when you're on the internet. Then you could go and use the internet, right? So like there was a serious gap of many years for people to get on boarded that doesn't exist in AI, right? Especially if Apple can figure that out and I just go hey Siri do whatever and it just does it. There's zero friction, you know? Um so it's it's there's we it's not a dot-com bust. It's not that doesn't exist here. >> It was it was Ben Bajarin from Creative Strategies who's been an advisor to all these companies for like 30, 40 years. So that guy that guy knows his stuff and I will say even listening to you guys talk I do agree on the bullish stance for Apple because that dot-com bust what you're describing Carl too is also it's like there were no devices that were easy to use and easily distributable at the time for people to get access to all this stuff, right? That was another huge problem is like Blackberries came out but man that was that was for like sending emails. It wasn't until iPhone, right? So even what you're even everything you guys are describing and you're talking about mom-and-pop shops and all that kind of stuff. It's like it's probably going to come in the same format of some next level iPhone that can just bring all that together for people. >> Why do you think meta is going so hard in the glasses and open AI is going to build whatever they're trying to build? They want to own the device, right? And it's hard to make people use a different device which is why Apple is kind of sitting there going, you know, at any point Apple can just turn this thing on and they're going to just change consumer agentic for everybody if they do it. It's a lot harder to get a new device in everyone's hands which is what meta is trying to do cuz that's their only option and what open AI is trying to do cuz that's their only option. So, >> You think Vision Pro was too early then, Kyle? >> Too expensive. >> Too expensive and too early. >> Too expensive and they didn't have enough use cases in it to just yeah. >> It wasn't subtle enough. That's why iPhone works so well cuz it's it's subtle. >> Want to stay ahead of the biggest technological [music] shift in history? Subscribe now to get insights straight from the sharpest minds in tech and finance. Quickly you'll note this show's for educational purposes only. [music] Nothing here is financial advice. Investing always carries risk. Never invest more than you can afford to lose. Thanks for tuning in. See you in the [music] next one.