Power & Memory Could Be the Best AI Trades Into 2027

There is one giant question hanging over this AI market. Are we building too much? The story is constantly that everyone is sold out of compute, memory, and energy and that we can't build this stuff fast enough. But does that mean we're going to end up with too much supply in the end? After all, that's happened to other industries in the past. But surely it can't happen here, right? What's up everybody? It's LGD here and welcome to Milk Road Stocks, the daily stock show that is here to question every possible take out there, including my own. Today is September 9th, 2026, recording way back on the 3rd. Every week, we sit down with Vincent, our leading AI researcher, as we dive deep into the economics of this bottleneck era. And this week, he's out here to challenge public perceptions where many headlines are claiming that this buildout is not needed and that we'll have more than enough compute with what we got right now. Those people are wrong, says Vincent. And the fundamentals are really starting to show it. We're going to talk robotics, Nvidia projections, and even Frontier model revenue as he crafts his thesis for the music to keep playing. This is usually the part of the show where I'll tell you you can check out Vincent's full portfolio for just a dollar in Milkro Pro at the link below. But instead, I'm going to tell you to like, subscribe, and drop a comment with your own hot take on this topic. Otherwise, enjoy the show. All right, Vincent, listen. There's a question on everybody's mind, and it's are we overbuilding AI capacity? And we've talked about this many times on the rollup. Like, is there too much? Right? by the time we get to building data centers on the moon or whatever, are we even going to need it? [laughter] It seems so excessive. Uh but I feel like you're going to you're going to help us answer that question today. >> Yeah, if you go to X to social media, you you you really get the sense that this AI infiltrate is dead. I actually think it's quite the opposite. I think we're in a consolidation phase, but the underlying data points of the last couple of weeks are really, really promising. and I want to run you guys through that today and show you where I think the attractive opportunities are. >> Perfect. Let's get into it. >> Yeah. So, let's start where we are and and and what kind of the fear is, right? So, we're building a lot of capacity, right? We're investing all those like the hyperscaler capex and and we're spending billions even trillions of of of dollars into that buildout. And I just added some screenshots here on the right that were from the last two weeks. I found like 15 or even 20 reports that had in their title, we're overbuilding, AI fears, bubble is going to burst, all that kind of to my view nonsense, but it's it's it's it's fueling the fear in the market, right? And you can really sense that investors are scared that we're building data centers that we don't need in a couple of years times and and and and there is no ROI on the invest right this is basically uh where we are today and [snorts] then if you look at the charts you can also see that and I found this one pretty compelling because the price toearnings ratio of the semiconductor trade is at its lowest since 2022, it's below the price earnings ratio of the S&P 500. While the earnings the earnings per share grew massively, right? If you look at in for instance at Nvidia, it grew between 30 to 40% a year um since 2022. So it's it's it's there is such a massive disconnect in what the fear of the investors are versus how the actual earnings of the company looks like right but hey we are in that situation and I think people need to understand that to to to also reckon why we're consolidating here at the moment. I just have a comment on those headlines you posted. Is that anything like that? I always think about how there's those headlines from 30 years ago saying the internet is just a passing fad. You know, like one of those one of those things which is like there will always be naysayers and people willing to pull it down. And you guys have mentioned this a lot too on the show that public sentiment is not good, right? That there is a very negative public sentiment and it's for the same reasons that there's bullish sentiment as well, right? that it's like well it's a lot of the same big billionaire names talking about this stuff the big companies people are mad that a lot of wealth is consolidated to the top but inevitably um I think that forms a huge bias for people where they can't really look at this stuff objectively right and there most people those people writing those headlines are not looking at any of these numbers [laughter] just looking at what they see in other headlines I'll let you continue >> I always try to find the the the steelman case right where could they be right or what am I missing in my analysis? Right? When the point where I end up is usually that the real the the people that make those bare cases, they do not use AI themselves. It's it's it's mostly the case because if you're a constant cla uh user, claw code user or or Grockbot user, right? you you you you're not using the the LLMs only but actually the agentic capabilities of those tools. You just have a different understanding of it. You just realize how much more efficient you get during the day like in every part of your life. Like me personally, I use AI easily between on a bad day maybe five to six hours on a on a good day 10 hours plus. I I'm only sitting in front of my laptop and and and and use the tools, right? And then you just get a a different sense of it, right? And if you if you go through the people who are making those bare arguments, I bet 60 to 70% of them are not using AI properly. They may be using CHBT and ask some questions about what can I cook tonight, but nothing more, right? And I I I I really do think that's the core issue. What if the consensus view is wrong and we're actually building or underbuilding AI capacity? What would that mean? Right? And let me go through a couple of slides on why I think this view is actually true and then in the end on what that would mean >> please. Yeah. If you look at data center construction for instance, you see that about 80% of the data centers that are currently being built in the US are pre- leased. So before their uh their construction is done, the money has been wired the the owner is already there and the data center will be used immediately and the the the spare capacity the the 20% that are left over are used within 6 months once a data center is built. Right? So this is telling you that there is just so much demand for for data centers that people are or or companies the hyperscalers the neoclouds etc are prepaying a lot of the of the capex to to to just give guarantees to those uh who who buy and build the data centers in the end. The question obviously is who who is buying those day or who is spending that money up front and why would they do that, right? And I I brought a chart here from the hyperscalers and the cl their cloud backlog which is above $2 trillion right now at the margin of somewhere in the mid30s. Right. >> Crazy. [laughter] >> It's crazy, right? >> Yeah. And and it and it explains why 80% plus of the data centers that are being built are immediately off the market or even before they're finished because the hyperscalers will take every data center is to meet that two trillion number of uh backlog because it's it's it's 30% plus margins on that. So why wouldn't you do that, right? So there's just crazy demand for for those data sets. >> 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 positions they're opening, it's just a dollar in Milkroad Pro at the link below. >> That is absolutely wild. Where I guess so, can we just go back to what you said last in the last slide? Actually, I'm curious about this. So, you're telling me that so so 80% is committed before completion on data centers, right? So, they're building it. So, basically, anybody's building a data center right now, and I actually don't even know who that is that's actually has one about to be complete, but when they do, like Colossus, I guess, is one that's that's been recently completed or about to be completed. 80% of what it can do is already committed. People have already bought that. Is that so how does that how does that how how do people call that how how why do people call that an overbuild? Like is it what we're saying that it's just like miseducation? Like people just don't the people running those headlines where we're saying like they just don't understand that. >> No, I think they I mean partially yes. I think people a lot of people are not looking at it's the same as oh data center using so much water, right? And and we're against data centers. It's such a nonsense argument if you look at the actual data. A data center I think needs a fraction of water versus a golf course for instance, right? So it just doesn't make sense and people are just believing on what they're hearing and and and reading, but they're not looking at the actual data. But that that that's one side of the argument. The other side of the argument, I think, is I have it on this slide here. people are not believing that this demand curve will be as strong in the future as it has been uh in the past. This is the this is the biggest issue here. But I think there's also against this argument there's so many underlying developments that speak against this or I brought three here with me. First, we had Nvidia earnings last week and a major major comment from from Jensen and Nvidia this week. They were meeting with one of the the big banks. I I don't know which Goldman Sachs or JP Morgan, one of them, and they told them, "Hey, on the earnings call, we we said we grew we we expect to grow about 70 70% in fiscal year 28, but we could grow demandwise over 100%. But we cannot communicate that that number because we're supply constraint right there's not enough power there's not enough memory etc to fulfill that >> 100% what 100% what grow 100% >> in revenue >> in revenue okay okay even despite despite posting record revenue they still Jensen is still saying we could actually double that in in demand okay got it they They the the the number they communicated on the earnings call was 70%. And they're sticking to 70 because 70% because they cannot meet cannot meet all of that demand because of supply constraints. There's not enough memory. There's not a power etc. Right. The street consensus the market uh was at 44%. So less than so they're they were off by more than 60% in terms of what how much Nvidia will grow next year, right? And this is just telling you how much the market is underestimating this this growth curve and and the demand for compute in the end. >> But hold on. But wait, let me play devil's advocate on that. Like that's just what Jensen is saying though, right? Like does he have any numbers to back that up? Like or is he just saying that? >> No, of course. Like what is that based? But but I say what is that based on other than just his data you like himself saying of course he has to say that you know what I mean like I would just say I just want to play the other side here too that it's like of course he has to come out of course as the street is saying it's going to be 44% demand and he's like no it's actually 100% like you know the worst thing in the world he could do right now is be like yeah it's actually the street's right it's actually less than that you know what I [laughter] mean even if I'm I'm not saying I'm not saying he's a liar. I'm just saying that that you know he has to remain bullish in some way. So the difference LG between a memecoin trader and the CEO of the largest company of this world [laughter] are actually SEC rules right and you [laughter] he has he's he's giving that number based on all the transparency he has today based on all the customer commitments he has into uh fiscal year 2028 so basically next year right that's where that number is coming from right if P were to be off if if that growth would only be 10% let's assume that the SEC would knock the door and would ask and and would ask some pretty nasty questions, right? And so so so publicly traded companies or CEOs of publicly traded companies can only communicate a number if they have a certain confidence behind that, right? Um that's that's what one of the biggest issues on why companies are not going public, right? because it's it's it's pretty tough to fuel sentiment, right? It's the issue that Tesla has at the moment, right? Elon cannot come out and tell us, hey, we're getting robo taxes this year for sure, right? Or we getting humanoids this year for sure. He cannot do that. He will go to jail if if if that would not be the case. So, um yeah, that that's where the number goes from actual customer commitments that he's having at Nvidia. >> [laughter] >> I know that, Vincent. Just you're telling me that nobody in the American financial system has ever lied about things. Anyways, I'm just saying I'm not saying I'm not I'm not I'm not insinuating that whatsoever. I'm just I just I just I'm just as we are discussing like this episode is about perception, right? Like this is what we're talking about today. So, it's like I'm going to come in with the devil's advocate uh you know, the opposite argument for a lot of this stuff when I see something that's a little off. Anyways, let let's continue. We're just talking about whether uh Jensen is a great dude or not. I think he's a great dude. So >> actually actually I think he's sandbagging the number like the number he communicated on the earnings call the 70%. I think he's sandbagging that. >> I think there's >> I think there's more more apps. So we'll come to that later when we're talking about like the procurement numbers of Nvidia. You see how much memory they secured until 2029. I think those are leaving alleviating some of the uh supply constraints that they're having. Um, but that is only a gut uh feeling here. Um, right. But yeah, it's basically >> you think that that's what he's seeing, but that he can't discuss basically. >> Yeah, I think it's it's that there's still some up upside in the market. Those are tactics from from seasuits when they're going out, right? It's it's always better to leave some room in terms of the number uh to the upside versus to the downside, right? uh to surprise the markets and and and >> yeah so but this is only Nvidia in terms of demand growth >> of course >> and that there were some actually from today I just had to to to to adapt this slide 15 minutes ago because figure AI the the the startup that is producing um the the humanoid robots in the US is buying um 100,000 GPUs where Ruben GPUs of Nvidia right and then you also have Tesla uh flooding Austin with all their cyber caps. And the message behind those two data points is we're moving away from digital AI into physically AI. And it's about to get rid. It's still early days for sure, but in Jensen's words or in Nvidia's word, they they posted this last week on on on X. We're saying physical AI will need 10x more compute than digital AI, right? So, this is the massive demand driver that the market is not um factoring in. If you actually go to the figure AI expost and listen to the CEO Bradco speaking to why they need the 100,000 GPUs, um you will realize that physically I requires so much more compute because it's a really messy world out there and robots, cars on the street, they need to be on point. They cannot make mistakes, right? And that's why they need so much compute that those companies can be sure they're putting out products that are safe uh to to society, right? It's just another massive demand driver. Um and and one of the core arguments in my view that we're actually underbuilding capacity versus overbuilding it. >> And see this is another this is another good part of the like public misconception probably, right? or the kind of like shallow analyst misconception is they don't realize that um what what optim or what Tesla's building with Optimus what figure is doing uh that FSD is coming and also we just did a show the other day with Melvin about kind of like the robotics buildout and you know he told me that Goldman had recently revised their estimate for total autonomous like uh humanoid robots release uh between now and 2035. It used to be 1.4 million and now it's like 6.5 or something like that. So they've like five or 6xed their projection of how many humanoids will be on the street like like released in the next 10 years. Um and and I think kind of like relating back to what you're saying it's like and if those require three times more compute then that's another reason it's like beyond data centers beyond you know asking chat what to make for dinner there's a whole other wave of needs coming and that's not even touching FSD right and he also said that you know Melvin also showed us that he he came he was comparing the amount of um I guess like autonomous devices or robotic devices that Amazon has versus employees right and Amazon has already had a million devices across their factories and it's not humanoids it's not people picking up packages but it's like arms and wheels and you know other little bits built across their factory. So they're already using robotics to a pretty extreme extent and as um as compute gets smarter as these models get smarter and they can train those a lot better you're going to see an explosion of that right and that's something that it's like I think that you know again back to those perception articles is like that's not factored in that that's what's being built out that that Tesla is pivoting almost their entire business to FSD and and humanoids. I think the what people are not understanding is that AI usage is not dependent on human beings eventually >> right >> today it is yes with LLMs >> but we're moving into a world with agents with humanoid robots with robot taxis those work without human beings right so the compute demand is not dependent like back in the internet days on human being usage and that is the the biggest unlock that makes me so confident to say that we're not overbuilding AI capacity here. >> Well, what does that all mean, Vincent? >> There are a lot of implications, but I want to highlight two here and and and and and then dive deeper on them subsequentially. I think on the one side this means frontier intelligence. So intelligence or AI from open AAI and entropic remains much more expensive than what we I think expected from kind of a commodit commoditization argument a couple of months ago. And and this is the I think the more interesting part from an investor perspective. the bottleneck trades that have the strongest scarcity, memory and power will remain very attractive. Um, and the earnings of those companies will go up more than what the market expects. So the the earnings durations will be stronger than what the market is modeling today. And I think they're they're great buys. But let's take that one by one. Let's start with the first uh argument on on on frontier intelligence. One thing people need to understand is if you listen to Dylan Patel of last week podcast, he's saying that 40 to 50% of new compute that is coming online next year will be bought by either OpenAI or Entropic. So you can say half of compute that is being built is is is used by Frontier Intelligence. Why is that? Well, a because they need that much compute to push for AGI to make sure they always have frontier intelligence, the best models available, right? But then on the other hand also because they are able to make that much revenue and I have a chart on that later and actually turn profitable which is expected this year. the an incremental gigawatt that is at at Frontier Intelligence or at OpenAI and Entropic pays for itself. So what I'm saying is OpenAI and Entropic um are able to pay more for a gigawatt of compute than other labs are because they know they can make more revenue out of it, right? And this is very bullish for the market because what will happen is they're they're they're bidding each other out. So they're driving the market price for compute up which is or which directly translates into pricing strength of memory of power etc. But at the same time it is the biggest risk factor in my opinion. It's not that we're overbuilding capacity, but it's actually that the frontier intelligence um or or openropic, right? It's it's it's a synonym are disappointing on their topline side. And if that were to happen, then we were running into a um over capacity scenario, right? Because they are the number one buyers by far in this entire trade. I think this is the biggest risk that people need to understand, right? This is the more nuanced view on it. >> 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. >> Does that mean that they are through a bidding war they overinflate the price? Like is that what you mean by that or like what do you mean by the topline risk? like what is what can you explain that a little bit more? >> Let's say the revenue of either OpenAI or Entropic does not even need to break down just stalls from here. Then they were a to struggle get more external capital and B they themselves would not be able to fund as much of the buildout, right? And this would kind of slow down the entire uh buildout, right? So, Antropic and Opener, I need to make sure that they're increasing their revenue constantly, right? That they remain on this growth path that they're on for the last couple of uh quarters. And maybe if we go to the next slide here, you can you can see it, right? The risk I'm describing here is non-existent at the moment, right? Those two companies are putting out numbers that have have never been seen, right? the the the growth is is is just exponential, right? But if that were to slow down and stall, that is the biggest risk because they're absorbing so much of the new compute that there would be a risk of over capacity. But that is the more nuance view that that that that people need to understand here. >> Got it. Okay. Okay. So So let me just recap all of that so I understand it is that they are how much how much of new compute are they going to buy up? 50%. >> Right? They're going to buy up so open. So the the frontier models are buying up 50% let's say of new compute demand or new compute next year, right? And they're out bidding each other for it, but they're doing that mainly because they can make more money per megawatt than everyone else who would bid on it, right? Because and and that's because now they're finally showing good revenue numbers. I feel like that re them showing those revenue numbers a couple weeks ago or whenever that was I feel like that was like a pretty big turning point in the market and maybe the market hasn't reacted to it but it seems like even doing the shows with you guys that that was like kind of a a very clear milestone when they were finally able to report like really decent revenue numbers and I forget what they were but that definitely seems like it's that has been a good pillar for people to build build out a a thesis for the next leg let's say. >> Yeah. actually were there were two different data points. First you had Entropic coming out showing their Q2 numbers which were 100% plus quarter overquarter growth and we were all like okay we're good. But then OpenAI came out and said we only had 18% quarter over quarterarter and immediately the market sold off because it was that risk that the market is fearing right that frontier AI is not growing as much but then we got the more nuanced view on on why why that was and how to interpret those data points um and and and I think now we're back on track. What will be really or what will be really important here is Antropic going public whenever it's supposed to be September or this year we don't know uh for sure but I think it will happen this year and then we have a lot more transparency in how those revenue numbers come together um how much of that is consumer growth how much of that is ad revenue etc um how how much each of those buckets are growing right then I I I I hope it this will this transparency uh that we get there will take a lot of pressure off the market because then the market finally sees okay we're growing we we're not overbuilding capacity right but yeah we're not there yet >> and again I feel like when those numbers came out it's it's still something that hasn't kind of come down into those even again going to bring it back to the perceptions you're talking about at the start it's like I don't think most people realize that that that's been one of the actual analyst fears is that these money these companies aren't going to make money in the end. They're not going to be able to show any kind of profit or revenue. Um, and now that's that's changed. What I mean, we've talked about comput a lot, Vincent. Um, one thing that you've been excellent at keeping track of for us is another major bottleneck, which is energy. And I feel like that's that's kind of the next part next thing you wanted to talk about because you've kept us up to date with things like Bloom Energy, which is definitely like one of your top like belief uh companies. But um I'll let you kind of explain what you want to say. >> Yeah. So if we take for granted now that we are actually under building capacity then the bottleneck trade remains on right and I think one of the most attractive expressions of that trade is still power. Two things I want to raise here. A Elon Musk came out this week was it last this week? last couple of days saying we have a power shortage of 15 GW or the other way around. We cannot turn on 15 GW of AI compute in 2027 [snorts] because we have a power ratio. Uh it's not only raw power, but there are not enough transformers and all and all the the different um input materials and products you need for the grid, for the for the for the gas turbines, etc. 15 gawatt. That's uh that's quite a lot, right? >> Yeah. >> But what this is telling you is that pricing power at those companies that deliver power regardless will remain crazy. And you can see it here in this uh Morgan Stanley research piece where they're saying we believe AI players will pay significant premier to ensure they are not short of power. Right? And this is directly translating into in the end earnings of companies like Bloom Energy who make sure a a data center can deliver power and and and and and or or or a data center gets power and and chips can be energized. And let me break that down a bit more. If you're building a data center, you're paying about 65% of capex to the chips and about 10% give or take for energy. Now, if you do not have power, you cannot energize your chips. So, what that means is you're willing to pay more or less any price for that power because you're paying 65% on the chips that are sitting there now and are doing absolutely nothing because you don't have power to energize them. Right? This is explaining the the time to power argument and why the margins of those behind the meter companies like Bloom Energies are or or will will keep going up because their pricing power is just strong. And it also explains why Elon Musk for instance with SpaceX comes out and builds their own uh or or bought actually a gas turbine company um or a piece of of of the supply chain there for actually to to produce the blades because those are the biggest bottlenecks in the gas turbine uh sector themselves because he knows um this is the biggest bottleneck and we need to focus on it. >> Yeah. Yeah. Of course. And I think he has been very vocal about that. And I think if anybody else wants to learn um a little bit more about Bloom Energy, specifically Vincent and kind of your your your um breakdown of it. I think we've done two episodes now so far about Power and Bloom. So you can kind of flip back in our channel. Uh look for any shows with Vincent and and there's a high chance they'll be about Bloom. And you've been calling that you called that in Milkroad Pro like months and months and months and months ago I think when it was sub $100. So it's interesting to see it recur. And I know you're a big bull on it, but it's interesting to for you to grab a lot of, you know, kind of new data points and new headlines that kind of, you know, even someone like Elon saying that's like time to power is so crucial, right? And that >> that that companies are going to be willing to to pay a premium for that, right? And I think that that's also really important, right? And even kind of relates back to to the open eye and anthropics of the world, right? Is that like they need to get this stuff online. They need to soak up supply and they need to get it online immediately and they're going to pay a lot for that. >> Yeah. And they're also going to pay a lot for memory. And this is uh something that Melvin has been on. Uh and I jumped on the train. Uh I I pretty much sniped the low. That was that was luck for sure. But uh memory is something I think that's not going away. And what I brought here with us or or with me is the Nvidia procurement chart of their uh earnings last week. [snorts] And what this is showing you um or or what you can see is that from October 2025 to July 2026, they increased their procurement obligations mainly for memory by uh more than 5x from 50 billion up to 279 billion. And those are uh commitments up until 2029. That's the first line on the chart. 88 billion mainly for memory pre-secured uh into 2029. Right? If you take that together with the news from Samsung this week that they um sold 70% of their uh capacity until 2031 with long-term agreements like Nvidia, Google, Microsoft, etc. This is telling you that the the the boom and bust cycles we had in memory may not be completely gone but are definitely alleviated and will be much softer. And that's why I think the the value of those companies, the multiples of those kind, I think you have you have Micron at a 6p, it's just ridiculous. It's it's it's it's deep bare market multiples, right? Are not true. And those companies are still due for a a repricing up. >> Mhm. Yeah. Absolutely. So, wait, so tell me that data point again. What's the PE on Micron? It's not five. That's not what you just said. >> I think it's six, right? Yeah. >> What? You can't be serious. I didn't know that. That seems insane. >> Yeah, it is. It's because what what the market is fearing here is it it has basically PTSD, right? Because what has happened in the past is that they had there was there there was a boom in demand. They overbuilt capacity and then because of that over over capacity, the prices fell and their margins fell with it, right? And we we saw that over the last couple of months, right? You you had micron Samsung as coming out telling us we're building new capacity to meet demand. But this time that the the the type of the agreements, the long-term agreements with floor prices built in are taking off that bust argument or or or issue that investors are having. Um and that's why I don't think this this this this market will run into into a bear scenario that what we have seen in the past preai. >> So there so so just to clarify that Micron's PE is 21 but their forward PE is like between 6 and 12. I didn't real I didn't realize you meant forward PE. What's the difference? Maybe just explain that for people like what's the difference between a PE and a and a forward PE. So PE is is is calculated with today's price to earnings and the forward looking is with the forwardlooking uh earnings >> where is that from their from their from their guidance basically like that from what they >> guidance and and exactly guidance and what uh investors are expecting basically right so that's why you can also have different forwardlooking PES >> yeah that's what I was going to say so yeah okay so it ranges bas and investors are what like like uh like is that like the goldman's and stuff like that of the world who are kind of suggesting like what it'll be. Yeah. Okay. Okay. That's Yeah. Right. So, their their PE is expected to drop that. >> Yeah. >> Yeah. Because they're expecting that their earnings will fall similar to what we have seen in the memory boom and bust cycles of of of the past, right? >> Right. And you're saying that that's not going to happen. >> No, because the data is Yeah. At least giving an indication into that direction. And I think the Nvidia chart is telling you that hey our primary constraint is memory and we are upping our supply or procurement commitments by more than 5x versus October last year to make sure we're not constrained by that bottleneck. So >> yeah. Yeah. Okay. I got it. And just just for clarification too that uh Micron also reports their earnings quite late every quarter. So, they're not due uh until September 30th, which I think will be another pretty uh significant day to look for, let's say, in the market. And and and we'll revisit uh these ideas here, Vincent, when we get there. So, Vincent, I guess the last question for you on this one is is how do we play this, right? Because this has mainly been a an episode, like I said, about conception, right, and perception rather, of what the market is doing, uh or rather whether this is going to be like overbuilt or not. You're showing us that it's not that there's demand and also showing us where that's from. So, what what how do investors play this? How do you approach something like this? Like you just while we're chopping, you just plow every dollar you have into it and wait for the market to realize. I have a pretty decent uh portfolio in the broader AI infra, right? So, I'm not uh putting in incremental or new dollars into the trade. I'm I'm shifting within the trade uh more into the bottlenecks I was outlining here. If you're not having a position yet, I think what Bloom Energy is 35% or even 40% below all-time highs, Micron is in a similar range. I think it's a great time to to buy the dip and and establish a position here and kind of DCA over the next couple of weeks and months into into those names because I think the trade will go on uh into 2027 and potentially 2028. Well, >> there you go. And you've said that a few times, Vincent, thank you for the episode, man. If you guys want to see Vincent's moves, learn more, and ask him some questions, all that is in Milk Road Pro, which you uh you can test out for just a dollar at the link below. Otherwise, Vincent, thank you so much, man. I I love shows where we talk about narrative. I'm a narrative driven person. Uh and I think that that's that's really powerful because those are the forces that play against people, right? Like when you're sitting there thinking, you're listening to our shows, you're looking at other analysts, you're saying like, "Listen, the fundamentals are getting better yet the price isn't reflecting. Like, what's wrong? Are they wrong? Am I wrong?" and then you have these headlines hammering you. Um, you know, that's that's that's typically a good opportunity, right? And I think it's I think these kind of episodes help help emphasize that. Thanks for listening to Milk Road. If you enjoyed the show, make sure you like and subscribe. [music] And if you're struggling to find winners in the market, that's exactly what Milkroad Pro is built for. Our analysts have called some of the biggest winners early, and Pro [music] lets you see what they're buying next, every trade they make, and the research behind every position. [music] Check out Milkroad Pro at the link below. Everything you hear on Milkroad is forformational purposes only. [music] 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.