This Is Why NeoClouds Could Still Explode Higher

Well guys, I guess my question for you is something that at least from my perspective I've heard you guys discuss a lot. We discussed this last year as well as well is kind of like what's the status of the hyperscalers and you guys kind of own uh a few a few of them kind of across your different holdings but is this something where even on this pullback you're looking to add are you looking more at the semi-rade instead? Like how would you guys kind of weigh those against each other? >> So let me just start there. I put up on the screen a uh a chart of the semis versus the hyperscalers. Over the last um few weeks they've really diverged from each other. The semis uh you know the hyperscalers just pulled back like crazy. Um I think we talked about it a bit last week where we said we weren't really buying a bunch just yet. Big reason why this is happening I think is because they continue to increase their capex. They're up to like you know some of them are like 20 to 40 billion a quarter they're spending. Uh and then they've stopped their buybacks. They used to buy back about like 15 billion at least Google did a quarter that's now down to zero. Same with Meta. And so, you know, if they're and now they're actually um looking to sell their own stock andor I think Amazon today announced they're going to do a $25 billion raise uh for their own bonds. So, they're taking out debt as well. So, the market doesn't love that necessarily. Um but I think they're starting to get into new territory where you want to be adding to them and buying because they're definitely benefiting massively from from AI. And um I don't think they're going to stop their capex. I don't know when the market actually like gives them the the green light again. Um but I do think um they're in good buy zones. Uh most of them uh Google and Amazon are the two favorites I have, but I haven't bought it yet. >> Do you think the meta news are bearish for the neoclouds and the hyperscalers? Basically, Meta is moving into the kind of cloud AWS business, selling excess computes, which I'm not buying into, but I want to hear your guys opin opinion on that. >> Melvin, I'll let you jump on this, but I I I don't think so at all. I think the fact that they're doing that tells you there's still so much demand in the market that they think it's a good business move, right? If it was the end of of compute and there was no demand left, they wouldn't make that business move. Um, and so I actually think it's it's probably bullish uh because it says that there's still tons of demand needed. Um, and they want to get a part of it. Whether they even do it or not, I don't know. But I think the key thing for these hyperscalers is like they might as well keep spending on capex because they either have the option of using that themselves to make their businesses better or renting it out to others who need it. That's why there's so much demand out there. So I I I think it's a bullish scenario for for the Neo clouds and for the the AI trade in general. Melvin, what do you >> think? Yeah, I completely agree. Um actually um I don't view this as a bearish news at all. I think market definitely overreacted. Um there's a lot of FUD that came out when the news came like news came out but a lot of people have you know counteracted the points like Morgan Stanley estimates that if Meta leases out just 250 megawatts of capacity to third parties at 40 $40 per megawatt it will generate approximately $3 in uh EPS upside in 2028. um which is roughly at 8% boost of their earnings. Scale that to 1,000 megawws, you're looking at additional $12 in additional EPS. So now, do I think Meta is going to do this? Absolutely not. Can they do it? Sure. Zuckerberg can do it, but I think this is more so a play from Zuckerberg to raise more capital to fund his capex because I know I I think Zuckerberg uh will likely raise his capex in the next coming quarter or the next few quarters. Um he he's one of the persons who's fully fully in this and this is this this is actually really good for Neoclouds because why? Because a couple months ago one gawatt of um to build out one gigawatt was about $50 billion. Okay. Now in just 3 4 month or within 3 months that price have jumped to 100 billion. So Meta essentially went and locked in all the capacity last year. They signed deals with Core. They signed deals with Nebius. Now the market in order for someone like an outside player to build this you're going to have to pay hundred billion dollars which no Neocloud or I mean only the major Neoclouds can really afford it and only the hyperscalers can. This is why I think Neoclouds will benefit greatly from this the the good the good players like the Nebias and the Coreies of the world because compute is only going to go higher from here and that there is that that's just the way it is. >> Yeah. I I think the metus it's a really strategic business move of suck because it's basically an arbitrage trade that he's doing. He's selling the capacity he has built for a premium. He's renting out NeoCloud uh capacity at a cheaper price and that's how he gets in the compute he needs to build out all his stuff, right? which we still don't know what he really wants to build with AI and [laughter] and on the other side on on the other side he saves some cost on on on on the compute right so I I like it >> I think the the market right now is just looking for reasons to be bearish and so and the market reacts on it too right this often happens in in bull markets you saw this like if you remember in 2021 if you were in the crypto space every time there was like any little bit of news on Bitcoin like some sort of negative news it would just rip down like 15% and then everyone go wait that doesn't actually matter and then it would just rip up again right and you can see the same thing happening here in AI it's just like you know you saw with Nvidia a couple days ago I think semi- analysis put out this thing saying that um one of the the I think next GPUs or whatever it was for Nvidia is going to be delayed like 12 months and the stock went down like% exactly and it wasn't even Yeah. It wasn't even true. Um, but the market kind of like tanked on that uh in different areas. Again, this morning you had DeepSeek coming out saying, "Hey, we're going to build our own chip and not use Nvidia anymore." Nvidia goes down like 3% in in pre-market. And it's like, guys, do you know how hard it is to build your own chip? Deepseek isn't just going to go start its own chip tomorrow and not rely on Nvidia. It's going to take them years if they even ever succeed, right? It's like it's just one report from routers or whoever did it and the market sells off in that. These are just great opportunities cuz I think, you know, as I said, there's still so much demand here. Computes going to continue to go up. We're going to need more and more of this stuff. And these are all just noise FUD that's happening in the market right now. Um, and you just got to kind of be able to to to look the other way, right? Not let that noise kind of get to you. Um, because these businesses are incredible. There's so much demand and it's not stopping anytime soon. >> We are at a very crucial point in time right now. AI stocks have ripped. They're going to keep ripping. Our analysts on the AI side are up like literally 100% or more on quite a few calls and crypto is about to boom again or at least it feels that way. If you want all the insight on what we're buying, what our analysts are doing, what's on their watch list, all of that is in Milkroad Pro. So join at the link below. >> I completely agree with you Kyle. The only thing the news on the Kyber architecture from semi analysis if this is true this is a really big data point because what that essentially means is that the entire 800 volt DC architecture is delayed by a year. So you need less power semis you need less materials. >> Yeah but it's not true. Nvidia already came out and said our road map's still intact. >> I I think they they ha they have to kind of deny it man. I I think they have to they even if it's true like they kind of have to do damage >> control. They just came out and said yeah yeah the road map is intact but there is no proof point from either side which is true or not. >> So um and and why would semi analysis come out and just announce something like that with no proof behind it. Right. >> They've done it before. >> They have definitely done it >> four or five times in the last couple years. So whether they're like trading on it I don't know but they've done it. Yeah, Dylan, the owner of semi analysis is getting like absolutely cooked on X right now because of like all the FUD that's going around. But I will say though, like I think this is a good like bullish sign for TPUs or custom chips, right? Like the Nvidia like the news that came out from Deep Seek today was a sign of that. Sure, they're probably not going to build it because like Kyle mentioned, the UV E UV tools, the memory that they need to build out all this is likely not going to happen. But there I will say this though there has been an increased number of growing list of AI companies seeking to like reduce dependence on Nvidia um by like making their own custom chips. Um like last month OpenAI un unveiled their chip jalapeno um with they developed with Broadcom um while Anthropic is now trying to build out chips with Samsung. Um and you might be wondering why is this happening? Because every AI, every hyperscaler in the world has the same problem. Nvidia is their landlord and the rent is too damn high right now. Like like think about think about it from like a like a hyperscaler perspective, right? You're spending hundreds of billions of dollars building out the AI infrastructure. You look at your materials and Nvidia is somewhere between 20 and 50% of the total cost and you're paying for you're not just paying for hardware once. You're locked into their supply ecosystem. every time they upgrade, you kind of have to upgrade because you want the best systems, right? So, like there's been a growing number of companies that are building it out like Google, TPUs, um Amazon with their tranium, and they really want to control their margins details because they have shareholders to worry about. Now, OpenAI and um Anthropic is likely not going to go through that because they want the best models, right? These hyperscalers maybe not don't want the maybe they don't want the best models. they they care about more about their price and margins, right? But what's interesting is Jensen is already making a counter move on this um last last week um like Jensen came out and he's basically fighting fire with fire. Uh we all know how much love Jensen loves Neoclouds and he's invested in like numerous um Neoclouds like the Nebus, the core weaves and now he's just going to use them uh against them to take on the hyperscalers, right? like uh what what what is he doing? He's he's basically Nvidia is starting to backs stop um starting to backs stop GPU financing. What that means is lenders can now fund Neocloud GPU clusters more easily because Nvidia is guaranteeing a minimum like revenue on the hardware. So this has been the huge huge problem for these Neoclouds. If you look at Corey who has like 50 billion $30 billion in debt and they have a market cap of like 50 billion, right? If you look at Nebius, they it's it's so hard to finance these because you have to like uh go and raise debt or raise equity, give up equity to do that. And Nvidia is basically saying here here's money, here's financing, go ahead and build out our systems and we'll we'll fund you. And Nvidia is going to take on a cut for that. And this is why one of the reasons why I remain extremely bullish on Neoclouds because Jensen wants a world where everyone thrives, not just the hyperscalers. and he will benefit greatly from that if all these neoclouds survives. >> He's their landlord and their bank now. >> Exactly. >> I find the timing of this really interesting because last week the CEO of Palunteer, Alex Karp, came out basically saying that the tokens of the labs are providing no value, actually providing risks to big companies if they're using those models. This week we see those Nvidia news you're just describing and Antropic and OpenAI coming out saying we give startups and smaller companies like cheaper APIs or actually compute for free more or less for f for a future revenue share. So the the the thing that I'm kind of trying to figure out is are the models really cooked? Are models just commodity? Because if they're giving out the tokens and the APIs more or less for free, yes, for a future revenue share, but if you're giving something to a server, you never know whether there will be a return or not. And what does this mean for hyperscalers in the end? If if their customers, OpenAI, Antropic, etc. are not having the revenues, can they actually pay the money to rent out the entire compute and how long can they do it? Right? That's kind of the risk I see in in in the broader value chain at the moment. [music] >> Want to stay ahead of the biggest technological shift in history? Subscribe now to get insights straight from the sharpest minds in technical [music] finance. Quickly, you'll note this show is for educational purposes only. Nothing here is financial advice. Investing always carries risk. Never invest more than you [music] can afford to lose. Thanks for tuning in. See you in the next one.