Why AI Memory Could Be the Best Trade of the Next 5 Years

What is the most obvious stock call that you have ever missed? And is the answer to that question for 2026, Micron? Many people on here love to cry about missing trades when stocks hit all-time high. But they are also the ones too scared to pull the trigger on those deep 30% pullbacks. And when it comes to the memory trade, retail loves to point at the cyclical nature of these stocks when painting the bare case. But it turns out this time might actually be different and higher highs still lay ahead. What's up everybody? Hey, it's Tjet here and welcome to Milk Road AI, the daily AI show that has the memory of a goldfish and has to ask the same five questions about the market on every single episode. Today is August 6, 2026, recording on the 5th. We have a guy at Milk Road whose portfolio is already up 50% since March. He called Micron at $300, Credo at $100, and many others that have ripped two to 3x. And today he's back on the pod to talk memory stocks and explain why the only way forward is up as many of these companies have already locked in their earnings for the next three to five years. Plus, he'll even give us a few lesserk known names to look at if you think Micron and Samsung are just a little too big for you. Melvin's entire portfolio is available to view in Milkro Pro, which costs just a dollar for a 7-day trial at the link below, and you can cancel anytime. A reminder that our podcast today is free and it wouldn't be possible without our partners at Securitize the Regulated Rails for tokenization and BitGet stocks 2.0 with real liquidity and real dividends. Keep an eye out later in the show for a message from them. Melvin, welcome back. Let's talk memory. >> Let's do it LG. Thank you for having me back on. Um, we did a memory podcast a couple weeks ago and I think it deserves another chance because I think memory is about to have a huge huge huge run up in the next um month or so. Um, and I want you guys to be ready for it. So, we're doing another podcast on memory. Um, so, but before we get started, um, LG, if you can show up, uh, throw up that first chart about the total addressable market. Um, so the total addressable market for memory chips was 214 in uh, 2025 and one year later it's 896 billion. By 2027 it's 1.34 trillion and by 2028 it's 1.68 trillion. That's a 8x increase in just 3 years and beyond that operating margins across the memory makers hit in are between 70 to 80%. For all three of the memory makers. So, this is a massive, massive opportunity and I think we're massively underestimating this and I I don't think you guys are ready for what's coming next. >> We got a little sneak peek of this a couple days ago. Martin did a similar he want to talk about SKH Heinik. So, if you guys want to deep dive into that, you can watch back that episode. But Melvin, I feel like you're going to give us all the similar but also different angle on this trade. But similar numbers, man. 77% operating margin at a 1.7 uh TAM is like berserk. like that's that's like that's absolutely there's like numbers no one's ever spoken about. So uh and and they're becoming reality in the next two years. >> Yeah. So we had a a huge pullback over the last like month month and a half. Like if you look at companies like Micron, SKH, Samsung, they all pulled back like 30 to 40%. The semiconductor index posted as worst monthly drop since 2022. And the nerd of you're hearing everywhere is that China is flooding the market uh flooding the memory market with cheap memory and the AI trade is peaking. The capex is rolling over. U the super sky is about to go away. But I'm here to make the argument that it's not. And this is a this is a long story and we're just getting started. So um that's why and and I have a bunch of trades for it. I'm going to give you guys like four to five names today. Um so you guys can all capitalize on um on these trades. four to five names. I feel like the only thing I remember in memory is three names. That's my LG's dropping jokes now. They're terrible. No more jokes on these shows. We're both banned from making show jokes. Anyways, okay. Tell but but I'm serious. Tell me um I guess let's let's start talking about those companies, right? Because there are some major players and like you said, maybe later you'll share some more obscure ones, but this is this is this market's pretty dominated by a few names man. >> Exactly. So, before I go any deeper, um let me just give you some context on memory. Memory is essentially the foundational layer of every AI system on Earth. What do I mean by this? Every NVIDIA GPU, every training run, every Chad GBT or you know every time you ask Claude a question, it requires all of that requires memory. You cannot literally build any of it without it. And three companies, if you look at the chart that LG just put up, three companies produce the memory that matters. It's Samsung, Skhinx, and Micron. And that's it. And these three control 89% of the glo market. And right now, all three of them are reporting numbers that would have sounded completely made up 2 years ago. So, let me just run through the numbers real quick because it's it's actually absurd. So, SKH Highix uh quarter 1 2026 revenue tripled year-over-year. First time in company's history crossing 50 trillion one. operating profit up five-fold in single quarter margins are hitting 72 75% and their HBM capacity sold out. Now if you go to the next slide Samsung look at that chart Samsung Q2 2026 operating profit 89.5 trillion one of 19 times where it was a year ago. >> What the is this chart man? >> Yes. >> I didn't even know Samsung was a player here. Maybe it's ignorant, but I'm like, Samsung, I got my TVs and Like, what what is this? This is like insane, man. How long has this company been around? Like 50 years. And these guys, they're making TVs and VCRs. And now now they're posting like a chart that looks like a meme coin. Like, it's insane. >> They've been around for a really long time, actually. And they have seen the Boston boom cycle, which we'll get into, but these guys are also like posting insane numbers. And then you look at Micron, who posted 41.6 billion in a single quarter. They beat Wall Street estimate by 16% and Wall Street already had a huge estimate on Micron and their operating profit margin is 81.2%.2%. And then their free cash flow is expected to go 30 billion next quarter. Now let me walk you through why the data says this is actually a buying opportunity and you should not be scared to, you know, buy these dips. So the old mental model for memory went like this. The demand goes up. The manufacturers build more fabs. Supply hit the market 18 months later. Prices collapse. Investors that invested in this lose money. And this repeats every single four years. And that cycle was brutal and predictable and and it the current market is not anything close or near any anywhere near that because let me start with let me explain why. because we have something called HPM now which is stands for high bandwidth memory. This is the specific type of memory that goes into the Nvidia GPUs or any any major you know GPUs like the TPUs and like other um uh GPUs that you can think of. This is this type of uh bandwidth like memory is physically bonded to the chip itself sitting like literally millimeters from where the compute happens. And that closeness is what allows a GPU to move data fast enough to run these actual AI models. Now, here's the crazy part. to make one uh gigabyte of HPM, you need roughly like three times the wafer capacity to take make one gigabyte of standard DRAM. So if you don't know what DRAM is, DRAM is basically the memory memory used in computers and servers to quickly store and access data. So So think about what happens every time one of these companies like shift a production line over to HPM to meet AI demand. They're not just adding HPM supply because at the same time they're pulling like the conventional DRAM supply out of the market like your chip uh for your phones or laptops and your PlayStation and that's why you see like a huge price increases for PlayStations across um Apple devices and more price increases are coming. So you actually what's happening is you actually get a shortage in both categories simultaneously, right? HPM is tight because the demand is outrunning the supply that the industry can actually physically produce and the conventional DRAM is tight because the capacity that used to make DRAM like the chips is used to make HPM now. And this is a completely different dynamic from any previous cycle because in the old cycle when high-end memory got expensive, the manufacturers just could pivot and flood the market with like cheap commodity DRAM to cool prices down. today doing that. You can't do that because it means you're abandoning HPM where they're earning like 70 to 80% like profit margins like we discussed to make to go to go back to making commodity chips at 20% margins that nobody's going to do that. They're not nobody wants to do that. If I was a company, if I own that company, I would not want to be doing that. And if if you look at the chart, you know, that LG has put up and the data conf confirms it because look at the chart of Goldman, right? Goldman Sachs, this is from June. This is a bit old, but it doesn't matter. It still applies. They revised significantly upward from their January forecast of DRAM prices up from 250 to 300% in 2026 with another 300 to 350% projected in 2027. >> 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. I need a telescope for this. Melvin will send you this chart in full resolution if you want. He'll redraw it. >> And then it's not just, you know, Goldman that's projecting this. Meanwhile, Jeff is calling for another 40 to 50%. Um Bernstein is projecting um like the average selling prices of gigabyte 2.23 per gigabyte by end of uh 2027. U nearly five times of what they were in 2025. And this is this is why I said this is not a boom and bust cycle because everyone is short on supply. No one has a you know supply anymore. So let me like show you like let me put aside the supply and let me show you the demand side because this is where the numbers are very hard to absorb because if you just go to the next slide over this is actually a good demonstration of um you know what the hyperscalers are spending. So if you look at the chart hypers hyperscalers spend 72.8 8 billion on memory in 2025. That goes to 335.7 in 2026 which is the which is a 361% increase in one year. Then 761.3 billion in 2027. So there is projections of capex coming around u from Morgan Stanley. Morgan Stanley projects that 1.3 1.4 4 trillion will be spent on capex and where is that most of that money going to most of that money will be directed towards memory because it's a commodity no one has a supply everyone wants memory to make these GPUs and therefore prices are going up this is why I am such a such a huge bull for uh memory at the moment to make this tangible a single like Nvidia black wall GB200 which are two two years old by the way require requires 192 GB of HBM3e. So one NVL72 which is like a whole cluster of rack that you know Meta and Microsoft have already deployed requires 13,824 GB of HPM and companies like Google, Meta, Anthropic, everyone's deploying this racks by thousands. So you can just imagine how much how much memory we're going to actually need because if you go to the next slide over this shows the current, you know, HPM supply in 2027 because this is a two the the the rack that I mentioned is a 2-year-old rack. Now now we're in like the Reuben production. Then we have Blackwells and or Blackwell then Reuben and then there's Famon that's coming in. These are all Nvidia chips by the way. And these requires even massive amounts of memory. And the demand for this is not stopping anytime soon because if you look at the chart, HPM demand goes from roughly 400,000 gigabytes in 2023 to nearly 5.8 million,000 gigabytes in by 2027. That's a 15x. >> From what though? From what? And I feel like we ask this all the time. You guys say it all the time, but from what? Like what what is this compute being used for? you know like we need all this comput we need all this memory but on what you know what I mean like what is driving the need for this memory >> yeah so the first few years ever since Chad GBT got released or before then we were in the training era where all these models are being trained and uh whatnot but now that models are you know there's still a lot of training happening now we're in the inference era inference basically means that every time you talk to a TAGBT every time you want to I don't know find something, ask a question or run your agents. That's what an inference is. It's collecting you know it's running your query or running your um you know chat and then giving you answers. That's what inference is and that requires memory. All the logs all the chat logs you have all that have to be stored somewhere and that requires memory. So this is why I say inference is causing a huge huge catalyst for memory and we're just getting we're just in the beginning era of this because agents is a huge huge part of this. It they're going to need massive massive amounts of memory because you you're not just having one agent. You have thousands of agent running in parallel at the same time. Then you have robotics that's coming in. Then you have autonomous vocals. All these things require memory and that is what what's causing these thousands of gigawatt gigabytes to go up in in like you know 15x in just like two years. >> Okay, Melvin, I have a question for you. This is this is an Nvidia chart and we know that Nvidia like obviously we did a roll up yesterday and guys were like listen Nvidia we we debated which hyperscaler and chip companies and everything to to bet on. Is this like a risk that Nvidia has so much demand for this? You know what I Like is it is it is this is it bad that there's one company that h soaks up so much of the demand? You know what I mean? Like is it is this too too concentrated? >> No, I don't think it's too concentrated because because it's not just Nvidia there there is competitors, right? Like you have AMD, you have a whole bunch of um uh like Google uh with their TPUs or Amazon with their traniums. they're also building out these you know uh GPUs or their version of GPUs. Uh but yes I I think it is a bit concentrated but it doesn't matter memory makers do not care who gets the supply or not. You know at the end of the day what they care about is their profit margins. It doesn't matter if Google spends a$10 billion or uh you know whether that's Google or Amazon whether that's Nvidia they don't care. Why is Google growing so much here in the next in the ne by by next year? Because they're growing as a huge share and I guess I guess so is wait who's the AMD as well but but Nvidia is kind of staying stagnant. Is that based on current orders basically? >> So Google's trying to catch up Google's basically trying to catch up to Nvidia with this demand. Okay. Yeah, because just to give you some context because the reason why these companies are building out their own TPUs and tranium chips and whatnot is is because there is Nvidia. If you look at like Nvidia chips, they're expensive. They're really expensive and one company controls almost all of the market. And if if somebody gets on a bad, you know, somebody crosses paths with Jensen and they don't if Jensen decides to I'm not going to give you any more chips, these companies are essentially screwed. Now Jensen may not do that, but you still don't want to bet all your chips on one company and they kind of want to break away from Nvidia and then make their own custom chips because they can also control their margins, >> right? Okay. Okay, that makes sense. So let's talk about like what when is supply coming right if you look at the next chart these are all these this maps every major fab expansion across all three memory memory companies and it tells you why s supply relief is not coming as soon as you you know as the bears need say it's coming so if you look at that chart micron Hiroshima fab is being built right now but HPM like wafer output doesn't arrive till like second half of 2028. Singapore is also um you know second half. Then you have Idaho meaningful volume maybe in mid 2027. New York is not till 2030. Um Samsung like they're they're not even coming online till 2028 2029. So basically what I'm trying to tell you is nothing meaningful is coming online before late 20 2028 at the earliest >> and what is what is a mega fab capacity like what is mega fab like what is this next wave of capacity like what is that >> there's certain number of wafers like a company can output like make these guys want to just >> make even more it's it's basically like terra fab in in a smaller level what Elon's building >> and Like let me just give you a caveat because even if these like fabs open supply it doesn't mean supply is going to follow right away because HBM is really really hard to make because it requires precision stacking tools that you know that takes 12 months to like manufacture and deliver. You cannot like you can have the facility ready like all these can be ready and wafer ready and still not be able to assemble because like because of the HPM like demand and there's a bunch of different equipments you need and it's it's a it's a really hard process and that's why there's only three suppliers in the market because you can't just like go spin up a mega fab or these these buildings and expect to make chip chips. It's extremely extremely hard to do. >> Mhm. Are they over are they overbuilding? >> No, they're not because >> there's no risk of that. Hasn't that happened in memory before? >> There there has. But let me just give you another stat because there right now there are 20 million uh GPUs um being deployed right now and that number is projected to go 15x in the next like 5 to 10 5 years or so. So all that all that need is going to need memory and like going back to the earlier point robots you have autonomous vocals Elon wants to put so many uh like uh chips and you know build out the data centers in space where's all that going to come from that is coming that has to you know some and all that need memory and Elon specifically said on this SpaceX um call yesterday that memory is the um is the bottleneck right now. So, this is not going to this is not going to be over supply because there's there's just so much supply that still needs to be needed. >> There's still there's just there's just so much expected buildout, right? And just so many plans to build that. So, unless that failed, then the memory will be needed. >> Exactly. On top of that, what these if you can just go to the next chart, let me just give you a counter argument against that because what these companies are doing because in every previous cycle, right, the same thing happened like I said, people just overbuild and the prices uh prices overbuild and then the demand just basically collapses. But what's happening now is hyperscalers are now signing long-term contracts with these memory manufacturers. Like I'm talking like multi-year like three to five years. And here's what Bernstein uh found when they modeled this. Approximately 35% of all memory bits shipped between 2027 and 2029, roughly 160,000 million gigabytes are already locked into these agreements at a floor price of 0.26 per gigabyte. And the current spot price is 0.29. So the the price is only 10% below where the you know where the prices are today. So even if there's over supply these these these memory makers have already signed these long-term contracts and they can just and these companies like that have signed these contract they can't just walk away because they sign contracts with something called the taker pay terms. this which is like essentially like a commercial lease, right? You sign the lease, you pay the rent, and the landlord does not care whether you your business had a slow quarter or not, you still going to get paid. >> Here's another question that's that's still kind of like a zooming out question and maybe this is, you know, you've probably said this many many times on other podcasts, but we have a lot of people that watch these shows that have never seen our shows before. So, it's always worth asking. The reason that these contracts are being locked in so long term, Melvin, is it to to scoop up supply? like is it for is it to lock in the price or is it to control supply or both like what is the what is the strategy for which to do this? >> So there's two sides. The first side is um hyperscalers like they need there is no memory left. So they need to actually go and lock down before somebody else locks down because if they don't they're screwed. They're not going to get memory to build out their GPUs, right? And that also creates a FOMO in the market, right? Like if your hyperscaler is signing 5year contract and you already locked in the supply for the next 3 to 5 years as a competitor, I need to also go and lock down these these things, right? Like it would be foolish for me to not do that. >> Yeah, of course. Yeah. Yeah. Yeah. And it breeds and it breeds that FOMO like you say just like as soon as one of them starts doing it and the others are like, "Well, we should do this in case in case even if we don't think that they're right, what if they are? We'll be screwed, right? We'll be left out on the D." So you have to do it. You have to do it. Yeah. And the flip side of this is memory makers. Like if you look at Samsung, SKH Highix and Micron, if you go back into the history and if you look at all the memory makers, there were a bunch of different memory makers. Where are they now? They didn't survive the boom and bust cycle. >> Only three have essentially survived all these times. So these guys are not stupid. They've seen this over and over and over again. And that's why they're willing to sign these LTAs now so they can protect their business in the long run. >> Right. And the idea too with these LTAs is that I guess I guess there's also a theory there which you kind of alluded to is that the LTAs would also prevent such a harsh downturn because it ensures supply for so long. Right. That's kind of what you said at the beginning, I think. >> Yeah. It ensures the price and they get >> ensures the price. Yeah. Yeah. Exactly. Ensures like you were saying your rent analogy that is like it ensures that they get paid. no matter what. So, as a result, um, regardless of if it's getting used or not or whatever's going on, it's like ensures they get they get paid. And in the past in past cycles, they didn't have these types of LTAs. I'm assuming there was nobody needed it for like we'll sign it for 5 years, you know, that didn't exist, right? They didn't do that. >> No, because everything was >> AI just changed the whole game. That that's the thing. That's the problem. >> Okay. >> So, so let me get into who benefits, right? How big can this get? If you just go to the next slide. So Namora data memory forecast um 60 billion they essentially forecast that just from the you know servers running AI 60 billion in 2024 107 in 20 107 billion 2025 and then 1.18 in 2029 and 1.4 trillion by 2030. So 23x in six six years from one end of the market alone. This is just from AI data centers and I'm not even counting all the other things that the these companies are involved in. So how do you benefit from this right? Which companies should you pick? I have some favorite my favorite one is always going to be Micron. This is the third. So this is third place which HBM market with approximately 20 to 21% of the market share and what's what I like about Micron is they're very strategic and also they are a US primary play because what I mean by this we all have seen the tariff war between China and you know other countries and whatnot the US especially this government is very focusing focused on American companies. How can we get American companies to dominate the market? And that's why I always always like to pick Micron because they're the they're the only American company that exists that can make this. And this is the same reason why I also like companies like Intel, right, and Dell because of the same same play. So what's cool about Micron is they are already signed like 16 strategic agreements. um these takeer pay contracts that I mentioned, hundred billions of minimum revenue is already locked in and customers have already pay them $22 billion upfront for the chips that they haven't even manufactured yet. Think think about how crazy that is. $22 billion have been just given to them for for stuff they that they haven't even produced before. 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 onchain. 9 years in, native tokenization, not wrapped, backed by Black Rockck, Morgan Stanley, and Kathy Woods Arc Invest, and chosen by the New York Stock Exchange, Van, 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. >> Everyone's tokenizing stocks these days, but almost nobody's doing it right. Thin liquidity, prices that drift from the real thing, dividends that just vanish. Bit stocks 2.0 is different. Real NASDAQ and New York Stock Exchange depth through licensed brokers. Prices mapped one to one. Dividends paid to your account in real time. Plus, you get the lowest fees in the market at just 0.04%. And you can trade them like any other crypto. as margin in earn in grid trading tokenized stocks finally done right. Head to milkroad.com/bitgget to get started. $22 billion advance to go and do your build your That's insane. And for anybody listening, if you were not if you haven't been in the milk streets up until now, Melvin, you called Micron back in like late February, right? And even that was late. That's And you only called it then because we didn't have a way for you to call it before that. That was when we launched our analyst portfolios. If we'd had this for a couple years, I feel like you would have called it a while ago. But even then, when you called it, like you were you were DCing at like, I don't know, 300, 320, 280, 350, like 400 before it went all the way to,200. Now, I think it's sitting around 900 or 950 or whatever it is today. But still, this is something where it's like you've been you've been part of this uh and and your bullcase for Micron still seems intact because like you're saying, I think that that's very compelling that that's like that's the main American company. So that is going to catch a bid and get propped up uh heavily to compete. >> Absolutely. Everyone Americans love their American companies. This administration particularly love American companies and Sanjay is a really good CEO as well. He has very good partnership with people like Elon and Nvidia and all the all the big CEOs. Um so I'm really bullish on this company. Now >> next company I want to focus on is >> Samsung. if you can just go to the next chart because I think this is one of the most interesting setups, value setups in the in the in the global equity markets right now. Uh, and I think it's undervalued compared to the rest because what do I mean by this? So, if you look at the chart, Korean equities are up 80 to 90% year to date. Um, and like a lot of people say like, "Oh, this must be a bubble because of it because you can't run up 80 to 90% in a year." But look at what the chart is showing. Like almost the entire return came from expectations going up and the multiples have essentially went up. You know, right now investors are paying less per dollar for forward earnings in Korea today than they were at the start of the year. That is not how bubbles work, guys. So this is their earnings are going berserk like I said before and this is why Koreans love or the market love these companies. So there's a lot of things that happened in the last like month or month or so that have really like m mainly four things that have amplified a huge sell-off. First one was foreign investors like you know like me I wouldn't call myself but like uh big players if you look at like the big foreign players they were piling into Korea since the beginning of like this year um like like your Samsung and SK Highix. So by June Korea's weight in global portfolios have risen above many funds active limits. So they had to trim because if if your portfolio goes up past a certain point, yes, you have to take profits and like rerate like all do all that bunch of stuff. So four investors sold a net net of 32.37 billion of Korean equity equities in back in June. So that is like the single largest outflow in Bank of Korea. So that was reason number one that caused this big sell-off in the Korean markets. So the second so because Samsung and SKH highix together represent nearly half of the entire Cosby index waiting any for selling you know in those names like the foreign selling like we we talked about just drags the entire market like entire market down and that causes a lot of mechanical selling with the algorithms as well. So that was the second reason. And the third is and we we know how much Koreans love to gamble. That that's their thing, man. If you look back >> that's Hold is that true? Do Koreans gamble more than other people in the world? Or did you guys just say that? >> No. No. I think at one point crypto like got banned in Korea because they know like they were about to gamble the out of this. I think it's might be still banned. Don't quote me on this. I could be completely wrong, but I heard that in a podcast somewhere. But >> people are This is the podcast where people are going to hear that now and they're going to be like, "I heard that in a podcast somewhere." And that was and it's just you just saying that and you heard it from a guy who heard it from another guy. Well, I got to look that I got to pause and look this up. Okay. No. Biggest most gambling countries in the world. Korea is not even on here, man. Okay. UK is number one. US number two. Australia, this is in order. Brazil. Germany. This is online gambling mainly. Italy. >> Okay. >> Canada. Yeah, we love it. We love that too, man. We love blown one hand. >> South Africa, India, Nigeria, Korea's not even here, man. >> I would I would love to see the data after this year because like dude, I'm telling you like 4% of all like Korean got margin called. Like imagine if like on the stock market. Yeah. >> Like which is insane. >> They gamble in their stock market. I think the figures I'm reading is like literally like online and casinos and like it's not it's not it's not leverage on stock markets. So I think I think we have to look for that separately cuz yeah otherwise it's it's it's I think it's mainly like our two countries that love that the most. >> Yeah. So the back to the third reason is because they created a single stock leverage ETFs um which created a feedback loop and that's what these Koreans were you know gambling their money on because they would just buy these ETFs and leverage them to the tea and just like lose bunch of money and then the fourth we you know we discussed this previously in the podcast before is because of situational awareness from Leopold fund um he was using forex leverage on skinex like the um stock and then that got forced to liquidate into Citadel during the crash that created a lot of force selling. So what I'm trying to say is I just wanted like all these are not structural things. These are all happen to like come at the same time. The prices the the the valuations of these are going to continue to go up. the the like the forward earnings on these stocks are really cheap and specifically for Samsung now I think the the bare case was for a long time they were like falling behind on SKH highix on HBM and qualification and SKHEX had better yields faster time faster qualification times and they were like the clear supplier for Nvidia but and the Samsung was viewed as the second right like the second place they were not going to get there. But that that all completely changed this year. Samsung passed the final HPM4 qualification with both Nvidia and AMD this year. They were actually the first ones to ship globally to Nvidia this year ahead of SKH Highix ahead of Micron. And their chips are freaking exceptional, man. like like Samsung HPM posted like faster data transfer than like like 11.7 GB was per second and Nvidia required like 80 to 10 GB. So they're putting in numbers. And what's funny is Samsung is not just a memory, you know, like company anymore. They have other businesses like the TVs and everything you mentioned. I'm not even I'm not even going to count that. They have like Broadcom partnership that spans like memory like leading edge chip fabrication at like 2 nanometers and below which is essentially what like TM subc do like if you don't know what TM subc TSMC do they make these chips right they manufacture these chips and Samsung has the capacity to do that and and interestingly enough there's reports that came out today that Anthropic is looking to buy uh make their own custom chips and who are they trying to make that through they're going to make it through Samsung. So >> yeah, totally >> that would be that would be a huge huge catalyst and that will have a rerating for Samsung. So and and Samsung's doing also doing massive massive amounts of buybacks as well which is actually good for you know stocks and dividends and you know for the stock price and whatnot. And that's why like I really like Samsung because they have not had the runup compared to Micron and SKH Highix. Pull up the charts and they have not had it yet. Well, they they have but not compare when you compare it to those two names. >> Do you have some? Are you do you hold Samsung in your portfolio? >> AS OF TODAY, YES. >> OH You have Samsung? >> I do. No. >> And that's a trade that Milk Road Pro members get to see. They would have been able to see your micron call. Do you have SKH? I do not have but maybe someone else in our in our company does. >> Yeah, they're gonna listen two days ago. They know who. But yeah. Okay. So So we got we got all of memory uh covered at Milk Road. Are there any name other names in memory that we should be interested in? >> Yes, absolutely. Since I'm feeling nice and I really want you guys to come join us for just a dollar, I will give you some trades on the equipment side because all these things needs more equipments and um uh supply chain needs to get rammed up. So, I'll give you a couple names that are that will benefit from this. The first name is Lamb Research. I think this is probably the most like direct bene beneficiary of of this. Um the company makes uh the company makes the machine to build and shape memory chips including the tools that etch like these tiny patterns into wafers and deposit materials needed to create each layer. So memory has grown from roughly like 39% of Lamb's systems revenue in in under a year in a year and they just raised 2026 wafer fab equipment industry to $140 billion. And the management what's interesting is management said something on the last call is demand is not the constraint anymore. It's capacity the clean room availability and like installation capacity that are the constraint. So this is a company that will massively benefit because of this memory boom. So like in simple terms like customers want more like LAM machines because they do not have enough factory or space or like workers available to install them fast enough. So this is like a critical critical path uh where it's equipment cannot be skipped. You need this to make memory >> and that's which company? >> Lamb research. >> Lamb research. Okay. >> Yes. Yes. Where does it trade? >> Where does it trade in in US markets? >> Yeah. >> Oh, Lamb Lam. Just don't type in with the B. All right, guys. I'm so dumb sometimes. Anyway, don't don't Google Lamb Lamb Research. Okay. It's going to give you It's going to give you a completely different kind of company. Uh, LM Research. Yeah, that's much better. >> That's right. And two other companies are Apply Materials and Kla Corp. Okay. So what these companies does is apply materials makes one of the machines that used to build chips layer by layer while KLA makes the inspection equipment that checks wafers for tiny defects before those mistakes ruin and expensive chips. So all these companies what I'm saying is the like you need these companies man to make memory and these couple companies that I mentioned are like the key beneficiaries of all this. You cannot skip these companies. Their volume and their demand is going up. Their um like their like profits are also going up. So this is an insane insane amount amount of demand. Um, there's a few companies I'm looking at as well, which I won't give out today, but um, if you're a pro member, you can definitely ask me for it. But, >> and they can quiz you and they can quiz you. That's another benefit of being in pro is you can go in the discord and you can you can ping Melvin at most hours of the day and he'll respond as well. >> Yeah. Yeah. Now, before I end all this, I just want to give you like some bare case because I don't think you you know I feel like I need there are bare cases, right, for for memory. So, >> and the risks too, Melvin. I think that that's the thing. I think I think generally like all this sounds nice and dandy and we see these charts on most of our episodes of like listen capex and all stuff and it all makes sense. It's it's it's sound theory and it's good even the weeks like last week you guys are like listen just just buy these dips man just keep buying and today we're seeing that payoff but I think generally like there there have to be real risks there is no bulletproof business case for growth you know what I mean there is no such thing there are there are risks and those risks are real so please you know you got to play devil's advocate on these >> absolutely so two arguments behind the selloff is CXMT which I'll explain what that is and the AI capex skepticism, right? So, China's largest DRAM maker just had their biggest IPO in like Chinese in China. Um they raised $9.8 billion. The stock went up like 466% a year or sorry on the day of the debut. Um market share market share has genuinely doubled for them for 4% to 7.8% of DRM in under one year. and they they have like insane amount of like revenue growth as well like similar to Micron and um everything but what I want to tell you is the big three still controls 89% combined and the CM CMTS gains are entirely in commodity DDR5 and LPDDR5 chips which basically are the memory chips that goes into phones and laptops like which is the chips that like Samsung and like SK Highness and Micron doesn't really care about it because they don't bring in as profit margin as you know compared to HBM the high bandwidth memory so they don't even care about this and to make matters worse there was you know we seen this talk between Apple and CXMT wanting to get some supply over the last like one month or so they were asking for approvals and what's funny is there's news that came out today saying that their price's price is actually higher than Microns and Samsung and SK Highix. >> They are charging more. >> They're charging more. Why is that? Because they don't actually like their tools. Okay. The tools that are used to make these they they like are they don't have tools tradition like compared to like um you know um Micron and SKH Highex and all these different companies because we actually stopped all these tools because um we don't want China to have access to these tools. So therefore, their output that they're making, a lot of it just goes to waste. So if a lot of it goes to waste, that means it's just going to come, you know, their margins, their their prices are just going to go up because they're making less of it. So this is like the the bare case around CXT, which I don't take them seriously. I don't think they don't even have enough capacity to fulfill their own needs cuz just think about China as a whole. They have their own AI domestic supply demand that are going berserk. They have so many freaking phones and like laptops over there and all that needs to be fulfilled by CM CM subt. So I'm not really worried about this. Yes, you'll you'll start to hear some stuff more stuff on this down the line, but it it's I'm not super worried. Next next thing is like AI capex skepticism because the problem right now is their hyperscalers are spending 7 like $30 billion on memory right out of like 1.3 trillion capex that's a problem like at some point me like all these hypers scales have no choice but to buy these but I feel like at some point they're going to look maybe into other alternatives Other alternatives being um we had the IPO of Cerebras, right? Cerebras basically like instead of cutting these wafers into like these smaller pieces, they just essentially make a big wafer, >> giant one. Yeah, >> exactly. So the all the memory and everything can sit together. Now that's I'm not hugely worried about it either because like you the constraint is still TM subc. You still got to go to TM subc for them to make these like big wafers, right? And and I feel like down the line there's definitely going to be like algorithmic improve improvements that we see like we already saw like the turbo quant like paper that came out that compresses like KV cache and whatnot. We had a sell off then but market was like reading that wrong. So there's always going to be these improvements in memory. That's just how technology works. Now if there is some massive massive huge you know changes or improvements in memory then I would be really worried. But as of right now, I don't see any any ways for them to really change the market. So, I'm not super super worried about memory. But like I do expect like short-term volatility. That's just part of these these stocks. But as we get closer to Micron's earnings, I think we're going to continue to like u make newer highs um down the line. So >> when when are the Micron earnings? >> Uh sometime this month. I forgot. I forgot the um >> My god, man. Micron's like your golden company. You don't even know how many the earnings >> are. That's because I'm not worried, man. I'm not worried. >> Yeah, that's fair. I think what I'm taking away from this episode is, you know, you've done a good job explaining to us like how each of the players work, where that demand is coming from. We had a lot early on about the difference between DRAM and HBM. And I think it's always helpful for people to understand that even if it's it's a little harder to wrap your head around and kind of the difference in demand for them. Um, but I think overall what I've learned from this show and also doing one with Martin the other day about SKH Heinix and um is not only do you everybody thinks that uh the Korean retailers or retail participants are extreme gamblers, but also that the cyclicality of memory has never gone through an event like this. And I think what really stood out to me today is the LTA part that you mentioned that it's like listen none of those cycles. Yes, memory is is cyclical because of like boom and bust cycles in tech, right? That tech builds and grows and whatever, but that nobody's ever signed these long-term deals and there's never been hyperscalers trying to just soak up as much of that as they can to the point where they're willing to sign long-term contracts that they never had to before. Right? So that even if there will be a cyclicality to memory and uh you know they they over supply or whatever that over supply is already accounted for. So as a result a lot of these companies they have business guaranteed for years to come as they continue to build out and also improve their product right and and improve um like you were saying with the me mega fab actually improve the technology. So if you know that's that's I think is is something that's very interesting that like regardless of what what continues to happen to these hyperscalers the demand has been locked >> these memory these memory companies are guaranteed these earnings for a long time right and that's something that in the the cycles of memory has never really happened >> exactly you described it really well that's a thesis in in entirety >> okay I have one last thing to share with you and it's it's it's it's on the lighter side so if you're if you're done learning you can you can tune out now but this I just I I asked Gemini. Gemini was the only credible source and it made me a chart of which markets like to gamble the most on the stock market. Uh so if you look at this this orange chart here, this is the average retail equity holding period in months. Uh and in the US, you guys love holding things for almost 40 months for over three years. That's that's great. Japan's two years, China is like half a year, and South Korea is like a month or two. So there's So you're right. You're right. they they love to just turn over their stocks. Um and and and let's see down here there's another chart. Percentage of volume driven by leveraged products. Uh so the US it's less than 5%. Japan is 10 is 5%. China is yeah they're pretty high. They're about between it's about 17 18% and South Korea is like the top of the chart. It's like 45%. So 45% of stock market volume is leverage in South Korea. I didn't know that. Man, this is this is actually I think this has been a big revelation for me uh about like markets is is how different markets work and why we see these. You know, I feel like following the stock market in in Canada, the US for a long time, you're like, "Ah, something's going on in Korea. I don't know what the hell's going on." You just hear about that once in a while and you're like, "It's messing up our man." As you're going to bed, you get like the ping and you're like, "Oh, no." But now, I think, you know, learning through you guys is like, "Wow, man. That's that's that's a market that that definitely likes to uh to play." >> Yes. I forgot to add one thing before before you guys go. I should have added this a little bit earlier. Um, so Samsung, if you actually want to buy the stock, you have to go through Interactive Brokers >> to buy the stock. So, just download that. It's a really easy process. You can buy it. Like, it it it's not that hard. But if you want like exposure to all three or all four like your SKH Highix, your D um your Samsung, your Micron, and even CXMT, just buy the DRAM ETF. And that has all of it and plus more. So >> just something to keep in mind. >> You need a referral code for Interactive Brokers man. >> I know, right? >> Interactive Brokers. Referral code Melvin Invests. >> All right. Thanks, Melvin. It's been a great show, man. I'll see you next week. >> Thanks, LG. >> Want to stay ahead of the biggest technological shift in history? Subscribe now to get insight straight from the sharpest minds in technical 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 can afford to lose. Thanks for tuning in.