SPEAKER_00 0:00

You are listening to the AI Investor Podcast from 24-7 Wall Street. On today's episode, we are going to look at just how ambitious SpaceX can be. What's going on with all this NVIDIA financing? Meanwhile, the memory versus optics debate continues to heat up. Intel stock, well, it's having a rough go. We're gonna look at Lumentum's earnings, and we're gonna do a rapid fire round. All that and more is next on the AI Investor Podcast. Eric, Earth's most ambitious entrepreneur, undeniably, is Elon Musk. And I would argue that SpaceX is now the expression of his ambition and maybe the most ambitious company on Earth. So you had shared a headline with me that SpaceX is looking to build out 10 gigawatts of capacity. Now, once we get to this scale, you know, this is sort of like going from a million to a billion, a billion to a trillion. These numbers get so large that it can be really hard to understand what it means to go, you know, to have a one gigawatt of capacity, which was an incredible milestone very recently, to 10xing to 10 gigawatts. So can you tell us what is going on here? And just how big is 10 gigawatts and what does that do to the AI space broadly? Is this inference? Is it training? Is it both? You know, to talk to me about what it means to have this much capacity potentially come online.

SPEAKER_01 1:20

Well, Austin, for context, right now, SpaceX and they've already received a lot of praise for how ambitious they've been in the past for getting pastors online very quickly, such as Colossus and Colossus 2. They have 1.4 gigawatts available. So when we're talking about 10 gigawatts, this is acceleration versus everything we've seen in the past of you know, sevenfold. I think today's podcast for listeners out there is probably going to be the most important one we've had in a long time. Maybe, maybe the most important one of the year, because what we're going to walk through is the economics of the next phase of this super cycle. And there's going to be some statistics that are really going to blow people's minds. But if you understand some of the underlying details, you're going to understand why SpaceX is move making the moves that's making, why Microsoft is making the moves it's about to make, why Amazon, why Google? You know, Austin, we've talked about it's the SpaceX race is kind of this um game of musical chairs, or you could use any number of analogies, but it's the fact that these companies kind of can't get off this race right now. And we're going to see some simply mind-blowing numbers for what's being planned for 2027, not across just SpaceX, but also Google, but also Microsoft, but also Amazon. And we need to explain the background why, because what this is going to do, this is just kind of raising the stakes for everyone. Um, what's being planned for next year, it's bigger than anyone would have forecast months ago, let alone a year ago. We need to talk about that. And we need to talk about what is the underlying factor and why it both raises the upside and the risk for this market. So let me just break down quickly SpaceX. We're going to take this in a number of directions here. I'll I'll try and keep it coherent. I can't promise I will. But yes, SpaceX, their ambition for next year, if you listen to the conference call, I think Musk threw out a number as big as 20 gigawatts. The realistic area we're looking at is six to 10 gigawatts. That would put them somewhere in the range of 300 to 500 billion in spend. As I already noted, that's an absolute number. But Austin, there's some research that came out on this Friday from semi-analysis. I know you, you, you, you can you can uh rasp me a little bit about my name check on them in each episode. But this this research was uh was so um kind of received across the market that SpaceX, which is a I think it's back to a $1.8 trillion company, it moved 15 or 16% on it, right? So you don't see companies at large move that much unless something has some data in there that's really impacting the way that people are viewing a company. And what they said is SpaceX actually has the potential to reach these levels. Last week we had talked about okay, SpaceX is talking about their earnings call this much. How are they gonna fund that? We're gonna need some more visibility, we're getting a little bit more visibility, but they also said that Google and AWS is gonna put them in the same tier. Austin, Google's spending about 200 billion this year. AWS is spending around 220 billion. This implies these two companies are looking at 50 growth next year, potentially more. Um, so Austin, the the key picture here is, and I I can pass it back to you because I know, I know we're gonna have a lot to unravel, talking about why these numbers are going up so much, why everyone is suddenly in this arms race that is escalating to a brand new level. Um, going to, you know, Austin, if if we were at uh a six before, which most people would say, what are you talking about? AI is at a 10. Well, if it's at a 10, we're going to a 14. Like this is the stakes headed into 2027. And we're going to discuss why it's bigger than anyone realizes. So I'll put back to you, but I think we've got a number of directions we're going to take this in.

SPEAKER_00 5:39

Yeah, I mean, this is in many ways, this is the biggest question in AI right now. And there's a lot of big questions. And as investors, you know, we're always looking for where's the next opportunity, the next um multiplication of consumption, the next bottleneck, whatever the case may be. But if we take a big step back, the thing that has underpinned all of this and this constantly accelerating RD ramp that we've seen for the last three years has been the belief that the demand for superintelligence is unlimited. And then therefore, this build out has a high ROI. So that has been so far proven out to be true in that you build up these facilities. And then, as we saw, SpaceX converted its own XAI data center to renting it for compute, and then Meta did a similar thing, and they have the ability to get a return on this investment relatively early. The other side of the argument that I'd love to hear you address, right? Not just to be, you know, we not just point at the stars and say it's going to be a smooth ride. The other side of the argument is that all of this build on this CapEx is being underpinned by inflated pricing because we haven't hit market saturation yet on the demand for compute. So one side the demand for superintelligence is unlimited, therefore we need to build, build, build. The other side is that all of the economics and the high ROI we're seeing from people being able to rent this capacity is still in the bottleneck phase and and a bull whip will happen eventually. I think I know where you fall on that, but I I want to, I'd love to hear you address that point because that really is the biggest assumption all of us. If pricing collapses on the rental for these data centers, it's that's the canary in the coal mine. And at that point, it's not even, I mean, at that point, the bomb might be going off. Um, so then if we agree with the first view that the demand for superintelligence is unlimited, therefore we can build, build, build, it comes to the second point, which is how is it going to be financed? And we've talked about how debt is now entering the arena, but the scale of what we're talking about will almost outstrip the capacity for even the hyperscalers to finance it without some version of seller financing. So I know there's a lot in there, but as you said, I do think this might be one of the most important episodes because the assumption that the economics of these data centers will continue to pay off is currently undergirded by the price that you can rent them at. So I'd love to hear you talk about that. Yeah.

SPEAKER_01 7:56

Yeah. And so I think there's two steps, right? There's the step of potentially getting to superintelligence and what that means. And then there's an intermediate step right now that we are finally seeing inference, the actual usage of uh models happening at scale, largely driven by the advances that we've seen in agents across the past mostly six months or so. So let's put some numbers behind it. Austin, this comes from the same semi-analysis piece, but as they calculate the neo-cloud average that's generally companies like um uh Nebbius, Core Weave, is $12 billion per year return on a gigawatt of compute. So, you know, if that gigawatt compute costs something like $60 billion, that's a 1x payoff over five years. That's that's kind of what you'd expect kind of a market to almost balance at. That's kind of the depreciation rate we have for GPUs. But but what we're seeing right now is the spot prices, what people are actually paying for compute to be significantly in excess of that. And that's what we need to kind of break down here. So B300s from NVIDIA, which is their newest system that we're seeing right now. They're gonna get to Vera Rubin soon, but right now you want a B300 system as spot price. You're getting about $29 billion annually. So over a four or I guess a five-year period, you're getting a 2.4x payoff. SpaceX's deal they struck with Anthropic, Anthropic's paying them $31 billion annually. That's a 2.6x payoff. And their deal with Google, they're getting $48 billion annually, which is over five years a 4x payoff. So, Austin, the question is why are these companies willing to pay so much? Why is Google essentially paying that much to SpaceX? And the reason that we're seeing is by um uh semi-analysis, own calculations. What OpenAI and Tropic are seeing from their own APIs right now is a gigawatt of compute is paying them about $100 billion annually. So that's a payoff of 8.3 times. So, Austin, I needed to put these numbers out there to talk about this moment and this arms race because the economics of inference are currently extremely attractive. And, you know, this is then a situation that's been in chronic undersupply, right? SpaceX, they are making a bet right now. They want to be first to serving this area that's been in chronic undersupply, which is leading to the pricing dynamics we're seeing. And they believe that if they can build faster than everyone, if they can get this compute stood up where they're renting out three to five gigawatts of capacity next year and keeping some internally for Grok, their payoff is going to be so fast that the question of how to finance it isn't as existential. They cannot build at this level if it's taking five years of payback, but they are not expecting five years of payback. Austin, a second consequence of this is a massive profit flip for open AI and anthropic that's happening right now. You know, the question is these companies are going to go public, right? We've seen so many companies in their growth phases go public being deeply unprofitable. And that's been the case for open AI. That's been the case for anthropic. But according to the analysis from semi-analysis, Opus 4.8 has an 85% gross margin for Anthropic and Austin. The other thing is too, the situation isn't just happening for these US labs. Deep Seek, there's a leaked investor call where they said they're getting a 10-month payback period on GPUs. Kimi K3, we talked about that recently in a couple episodes. Once they released their model, they essentially had to shut off inference because they were getting hammered so much. They just couldn't serve it. And what we're gonna see in the near future, because of this market dynamic, Anthropic is now targeting an IPO in the next 60 days. So if you have a massive acceleration ramp in inference, it's it's likely going to IPO Austin at about probably $2 trillion because they're going to be showing that they're extremely profitable, they're scaling extremely quickly, and they're definitely underpriced in their recent round. So, Austin, the combination of SpaceX, Anthropic, OpenAI, we're probably looking at them taking advantage of this moment of this spot pricing of their placement sent to all IPO in a six to nine month window. And right now that looks like where they're IPO and it's gonna land and putting them value at around $6 trillion. That's equal to the market size of the UK, Brazil, and Italy combined for just three companies. So, Austin, you know, I do have a few things here. I'm gonna come back to this longer term picture. Um, but I do think, you know, there's there's some dynamics at play I want to talk about, um, how this is going to impact not just SpaceX, but also secondary companies like Microsoft. Uh, what's what's happening at Google right now? And then I'll take it back and close it up. But I did want to give you a chance if if there's anything, you know, from there um you wanted to double-click on, or, or we can just marvel in amazement at what how powerful this window is right now.

SPEAKER_00 13:34

Somewhere out there, the Guinness Book of World Records is just loving this year, right? But for all the records that keep getting smashed. Maybe these are only the records that we and our listeners care about, but the scale of these is truly amazing. I mean, to think that we had the SpaceX IPO, uh, you know, biggest IPO of all time, and now we are going to quickly follow that with an IPO that's probably 2 trillion, uh, is it's truly an amazing time to be alive. Speaking of Google and Microsoft, which you just uh mentioned, one of the things I also want to remind investors, you know, that the bare side of this argument, which is that this pricing that's underpinning this data center ROI, right? You talked about a five-year payback. At those sort of economics, you build, build, build, build, baby, build, as the uh expression goes, but that is still underpinned under these elevated spot rates that um these chips are getting. So the argument that maybe that pricing will collapse, and then therefore this building, the build-out might collapse. One of the other things to remember is that we saw Alphabet and Microsoft adjust their own depreciation schedule on chips from three years to six years. And they came out, I believe it was Google came out with a lot of data and said, look, these old chips that you thought only had a three-year useful life, we just downgrade them to lower level tasks like YouTube and Gmail now, which is another way of saying all of this capacity that's being built out. One of the assumptions that's underpinning it, not only that price will remain a high, but implied in that is that the only use of this, uh, these chips will be AI training and inference. And once we get three or four years into the future, there's new chip cycles. Those older chips could be used for other web processing or lower level inference as well. So there's this whole like long tail industry of commute, uh, compute demand that Google has shown us exists way below way beyond what people were expecting as well, which further gives credibility to the idea that you can afford to build out these massive data centers and see you know an ROI for years to come.

SPEAKER_01 15:20

Yeah, and the other side to that, too. Just this morning, we're recording this Wednesday, August 12th. Uh, Core Weave uh reported their earnings, or maybe it was last night. I everything's blurring for me right now. But the point is here, they they as part of that, they talked about they had signed a new agreement for A100 chips, taking them into 2029. Austin, A100 chips are a 2020 product. They're ancient. Yeah, they're ancient. So so that's that's they are now it's it's a nine-year lifespan for those chips, which you know, it is just again showing some of the positive um pricing dynamics that that continue building enthusiasm throughout this market. But I did want to talk about Microsoft here because this this was another part of the semi-analysis um article, how how they're going to really push for a unique moment, which kind of stands in contrast, you know, Microsoft after their earnings, they jumped 15%. And part of it was being applauded for their kind of safe approach to AI relative to many of the other companies. The reality is they're planning a massive ramp. Um, they have a little bit of an air gap that's happening right now because they had taken their foot off the pedal, but they they themselves, Austin, have 10 gigawatts contracted, which is, you know, the same size as SpaceX. And last quarter, they started taking on a massive amount of third-party leasing, which is, you know, inventory from NeoClouds. They're gonna be a prime candidate to take on some of SpaceX's build out itself. And Austin, this is because of a unique advantage they negotiated, which is they have access to OpenAI's models served by Azure and for OpenAI's applications. So, Austin, what's happening right now, and this, I need to talk about this to explain why we're kind of in the situation of everyone pushing as hard as they can at this moment. If Microsoft can get the compute, they don't need to pay this. I think it was a 20% tax that they had. So, what they can do, according to semi-analysis estimates, per every gigawatt that they're getting of these B300 systems, they can get $100 billion annually. Let's put this in context. Azure's current run rate is $124 billion. So if they can, Austin, if they can get several new gigawatts on next year at this B300 equivalent uh rate of serving tokens, we could see absolutely massive acceleration. And if they have this acceleration, it's going to make it far easier for CEOs to come on their calls and talk about the ROI they're getting. Amazon, after their earnings, they took their CapEx up to $220 billion this year on their earnings. But Andy Jassy did a full-throated defense of the ROI that they're seeing, which is kind of what Wall Street wants to hear. It wants to hear kind of what these companies are doing. And, you know, Austin, another side of this too, this the past week that we need to talk about is what's going on with a company like Google. I've talked a lot about Microsoft's prisoners' dilemma, where they can make so much money renting out their compute, but they also need their compute to defend their core products. Google has a version of this right now. And what we're seeing is the founder of DeepMind, he's out. He's out amongst what's being called a talent exodus. Jeff Dean, who's been called the godfather of AI at Google, he was responsible for a lot of things like the creation of TPUs themselves. He's leaving the company to go for his own founding. And what we're seeing is Google was seeing an 82% cloud growth. Well, Austin, that's because they're giving their capacity to rent it out, to take advantage of these dynamics in pricing right now. But the end result is they're falling off the frontier. They don't have a competitive model right now. So what Wall Street applied in the short term has severe long-term consequences. And you're seeing a divergence right now, Alphabet down 3% across the past month. Microsoft up 29%. But Austin, why I wanted to talk about this is to put it all together a little bit that we are going to talk about risk. Actually, let's just talk about risk right now. The big thing is there's a lot of companies with unique moments that's causing them to push harder than they otherwise would. SpaceX, you have this spot market. They have a deal with Google where they're essentially getting close to paying off their GPUs that they're renting in a year. You have a lot of robust spot markets. You have companies like OpenAI and Anthropic willing to pay through the nose right now because their gross margins are so good. The more compute they get, the more they're going to be able to sell it at outstanding gross margins with the current dynamics. And then you have a Microsoft, the one company that you'll hear people say, oh, well, you know, they're they're the most conservative of these. They have their own unique dynamics that they get access to open AI's model, and they can potentially get at current rates $100 billion per year by serving OpenAI's models, which is going to make them want to push as hard as possible. So you've got all of these companies with a lot of self-interest to push as quickly as possible. But Austin, the big question is if we are moving towards an area like superintelligence, where you have a market for intelligence that's worth $5 trillion, maybe $10 trillion annually, are we going to see an extreme kind of rebalancing in the near term? Because actions that made sense at this particular moment and made sense to each individual player led to such a supply overglut that we had to basically reset the market. I am a little bit worried about this. And I wanted to break this down to show everyone SpaceX is going to be putting. Pushing harder than just about anyone forecasted weeks ago, let alone still today. Microsoft, they're the one people are talking about taking their foot off the pedal. They're going to have an extreme incentive to push as hard as possible and is going to look great in some near-term results. They are going to put up some eye water and growth rates for Azure. Google is behind. They can't get their frontier models competitive. They need more compute. They have an interest in doing this. Anthropic, OpenAI, they're getting such good economics. They will buy whatever compute they can get. Amazon is going to spend $300 billion plus next year. It would certainly appear. But again, Austin, it is so much, so quickly. We've talked about we we would rather almost this was more of a gradual, a gradual climb, but the dynamics to each individual company call for all of them being as aggressive as possible. If we have the demand for it, this is going to work out to a situation that this market structurally increases. It's going to grow into that, you know, $2 trillion a year spend level that we've talked about being a current ceiling. The other limiter, Austin, is all of these companies have the ambition, but will we potentially be saved by some near-term pain by the fact that there's only so much capacity from Taiwan semiconductor? There's only so much capacity to build the components.

SPEAKER_00 23:36

That's the self-regulating metric in the market right now. People want to go. I mean, the the self-regulating forces right now are Taiwan semizone capacity, memory bottlenecks and capacity there, interest rates, right, could could be a retarding force on all of this. The difficulty and challenge of all the debt. So we have said we want this to go slower, but maybe we should also appreciate there are a couple sort of built-in market regulators already smoothing this a little bit and tamping down ambitions. You said you're a little worried about this, all this capacity coming online, though. What would you watch and what do you recommend our listeners watch to see that story pay out? Is it like, is it H100 spot pricing? Because I'm pretty sure H100s are still pricing where they were in 2023. And I think the B200s continue to increase. They're up about, you know, up around five or six bucks now. So is it spot pricing? Is it capex from the hyperscalers? Is it debt financing rates? Like, what are the key metrics you would see to say like, okay, things are things are cracking, things are softening here, or no, this the thesis is still intact? Is it all of those things?

SPEAKER_01 24:46

Yeah, I you know, the thing about spot pricing, I don't want to get too granular here, but there was a period a month or two ago where it was coming down, and everyone started saying, ha, see. Um the thing is, this is a market where you're going to see some waves, and you know, it's more important to sometimes watch things like what value this is being contracted out at by companies versus, you know, any spot pricing necessarily. But we're gonna need to watch that more, Austin. We're gonna we're gonna need to see what's being paid per unit of compute headed into the future. It's that's not something that we've always talked about. It's probably going to be more important. We're probably going to need to spend a little bit more time to your point about Taiwan semiconductor, seeing, you know, what the rate limiter of some of these initial inputs. And it's pretty interesting though. SpaceX, how are they going to do this, right? How are they possibly going to build six to 10 gigawatts if all that we hear about is power permitting doesn't allow for new construction? If all that we hear about is G Vernova's jet engines that you know are using for this behind-the-meter power are sold out through 2030. How are we going to do this if all we hear about is how limited labor is? And and it's interesting, what what they're essentially doing is one by one, they're just finding alternatives for major bottlenecks. So they're going to skip switch gear and transformers that have multiple-year bottlenecks by looking increasingly to power modules from China. Um, you know, they're gonna skip turbines from GE Vernova, and they're gonna look for secondary sources for labor. When I went, I visited um Stargate Abline. I think the number I had heard from one of the workers was 12,000 people coming onto the site daily. And to build Colossus 2, they had about 3,000 peak laborers. So um they're they're just going to find some alternatives. Now, I do want you in the middle of this, Austin. I'm gonna bury it. Sometimes when we look at when we look at our watch charts and our view charts, we see people they look for the recommendation and then it spikes and then it goes back down. And you know, if someone wants to do that, it's your prerogative. But I I think just grabbing a recommendation without it, it's often so much more, Austin, right? It's not that specific stock. It's just think about the opportunity. And and you know, I've always said I'm I'm more concerned about trends necessarily than stock picks. And and I think that is going to contribute more returns. So I'm not even gonna put this buy recommendation in our show notes. This is just gonna be part of this section. This is an Easter egg sneaky.

SPEAKER_00 27:26

This is an Easter egg for the real fans out there, right? You know, Jorge, we know you're listening. Nolan, we know you're listening. These are for you guys. This is for you who show up and listen to the whole episode, not the people who are, you know, scrub scrubbing just to get the ticker symbol. Um, you know, I to just to close out this SpaceX point. Sure. The this is Elon Musk's zone, right? What he does is he builds. And he obviously, you know, he is uh he's a software developer. He did it with Zip2 and PayPal. And he also, I mean, he's very good at software too. Tesla self-driving is remarkable. They've reworked the X algorithm. But when you actually think about his real, real, most core, unbelievable, unrivaled strength, it is this it's manufacturing, it's building facilities, it's working through the physical um atoms bottlenecks, not necessarily the bits. You know, Grok is not the most leading frontier model, but nobody else was able to string together 20,000 NVIDIA chips in a coherent manner. And Jensen Wong, you know, the maker of these chips said that was incredible, and only Elon could have done it. So when you talk about 10 gigawatts of capacity and how SpaceX is working around these bottlenecks and skipping turbines from Vernova, like Elon just operates in his own ecosystem here. Like, right, it's the reality distortion field. Other bottlenecks don't apply to him because you know he understands physics at such a fundamental level that he can work around these things. Look at what they did with the Tesla model three, you know, manufacturing under a tent, right? Of going through production hell, scaling the supercharger network. This is what he does. So I am not, I say that because it seems inevitable to me that SpaceX will build 10 gigawatts of capacity. Other people, other builders might run in, or I expect will run into challenges that he will not, or they will not meet the pace that he does. So if for the foreseeable future, let's say for the next 24, 36 months, spot pricing holds up and the demands for compute holds up, SpaceX is the company you've got to pay attention to because they're gonna be bringing the most capacity online, the fastest, and working around all these shortcomings. It's actually the people who bring capacity on later where you might see those challenges that you were saying, like, yeah, I might be worried about this, right? If we start getting 20, 30, 40 gigawatts coming online. It's the later uh uh data centers that coming online where things might become an issue. But incredibly ambition. I want to see, I want to see how they do this. I love hearing about how they work through these shortcomings, like skipping turbines from G Vernova. It's just it's amazing to me. The scale and the pace is truly amazing.

SPEAKER_01 29:57

Yeah, you know, it's it's interesting too, because you could just take, you could count up. You could say, all right, SpaceX 300 billion, Amazon maybe 350 billion, Microsoft 275. You could count that way. The problem is you're gonna be double counting some there because that SpaceX capacity is probably gonna be going out to those other companies you're talking about. And because SpaceX will be so aggressive about building fast, they're gonna probably be looking to some outs in other arrangements they found. So if SpaceX can build fast enough, um, you know, it's going to replace some of the other third-party uh build outs and it potentially replaces some of the directed build-outs that these companies are doing. But Austin, I'd I'll just say quick the company I want to recommend today is Solaris Energy Infrastructure. It had sold off big time in the situational awareness drawdown. It was one of their top 10 positions. And in short, what they do is they provide a fleet of gas turbines and infrastructure for companies to quickly set up behind the meter. What you like about them is they have an existing relationship with SpaceX. And they're a really logical beneficiary of this news across the past week. When I look at their valuation, I think it's pretty reasonable in terms of what kind of earnings they could achieve by 2028. And they're still trading about 30% below where they were at recent highs. So, you know, you're not buying them on the bottom, but you're not paying the maximum price we saw when, you know, the fervor for AI stocks really hit its peak in early June. And Austin, they're they're still very early. Um, they've got decade power at scale for two data centers. They've got two more coming in September, and they've got long-term contracts with three leading technology companies. And this is all developed in the past six months. So I like this stock. Um, it's just one that's got a really clear catalyst throughout this 2027 build out we're talking about. You know, we're talking about there's going to be this build-out. It's going to be rabid. All these companies have incentives. Everyone's underestimating it. That's why we want to talk about because these unique incentives for everyone to push hard. And this is a company that's going to, in the background, see one of the clearest catalysts from it. So I think $10,000 when we're thinking about how much we want to put into it, $10,000. And it gives us more of a space in this behind-the-meter trend. I've talked a lot about one company that's not an official recommendation of the portfolio, but Austin, I've talked about so many times on the podcast, is Bloom Energy. That stock's up 15% today. And that's because Nebius is moving um power in a New Jersey site to Bloom. Um, Bloom remains my largest behind the meter bet personally. And I would say related to, as I I quickly talk about behind the meter, one stock we get a lot of questions about is Power Solutions, PSIX. We had recommended it last year. You know, it was a time when, you know, things were feeling a little frothier at that moment. And we had framed it as speculative and risky. It has been just that, but I I think a lot of people bought it. So I get questions about it. I've said I've been on the cusp of selling, and I haven't officially done so. And it's because of the situation we saw in the past week that this company might be a poor opera, they might have poor governance. But why we bought them was because we thought the behind-the-meter build-out was going to be so intense. A lot of these companies were going to just be looking for any supply they could get their hands on. And this was a company investing in supply growth that would be able to meet that. This is what we've seen from SpaceX. SpaceX is going to look for companies like Power Solutions. This is who they're going to go to to be able to find this incremental inventory that allows them to beat everyone else to market and get around these choke points. So this has stock rallied 22% the past week. And this is exactly why, Austin, because there is so much ambitious, um, you know, ambitious build out behind SpaceX. So this is, yes, there are some risks to how quickly this whole situation is. We're going to see more growth next year than just about anyone's factoring in. Uh, you know, maybe people are starring, it's starting to dawn on them, but there are some companies that are going to get you ahead with re-rating on this situation, um, which, you know, is what we want to capture. So, Austin, you know, I I don't know, you know, how to close this per se. It is kind of tough to say, hey, I I guess what I'm trying to tell everyone that's listening to this podcast is there's going to be a lot of headlines. There's going to be a lot of profit growth that's bigger than anyone realizes. That's going to be very good in the near term. We do want to balance it with the fact that the uniqueness of this environment of everyone trying to cash in as much as possible on these extremely profitable inference dynamics will create its own risk. So I don't think that changes anything in the near term. And in the long term, frankly, if we have a lot more capacity and that that cause some dislocation in infrastructure stocks, on the other side, we're hardened to see that, as we mentioned the last episode, there's beginning to be a lot more of an ecosystem of companies being built on top AI, things like Palantir gain momentum, things like our recent recommendation and service. Now, companies that are going to facilitate the growth of agentic AI. And if we're able to continue across the next year, diversifying more into that versus just infrastructure. Well, I think that's a pretty good thing, right? So I think as of right now, this causes no change in how we're investing, but it will cause a change in how we continue evaluating the market going forward.

SPEAKER_00 36:04

I I really like this recommendation for Solaris, Solaris, I'm not sure how to pronounce it, because this feels like a rationalization. Sorry, the the rational bet, the rational version of the lazy bet that people made, you know, a year ago and 18 months ago. And the analogy that is similar that comes to mind for me is nuclear. So everybody was seeing all of the power needs that these data centers would need and just the sheer scale of and saying nuclear's got to be the answer. So a lot of nuclear stocks went, forgive me, nuclear, and people were just bidding them up on the amount of power you can generate without any sort of thought to how long does the permitting take? How complicated is it? Data centers already have local regulatory pushback. And when you when you throw a nuclear reactor on top of that, you you get a heck of a lot more. Um, and your permitting uh bottlenecks become even more damning. So Solaris Energy was a part of that like turbine GE Vernova catch up pick where everybody was just like run scrambling on top of themselves to say, okay, you know, we want to buy nuclear, let's bid them all up. Okay, we need turbines, we need natural gas, let's bid them all up. Now we're getting this little bit of a pullback, probably by the situational awareness moment as you talked about. But this one is specifically tied to SpaceX, which I really like. So this is not sort of like the caveman, you know, brute force, swing a club, let's just buy natural gas, let's just buy turbines, let's just bid up um nuclear. This specific company has direct exposure to the most ambitious, fastest moving company in the space that is working around the industry-wide challenge of a GE Vernova supply constraint. So I really like this pick because it's sort of it's it's the smart, it's the it strikes me as sort of the smarter play in the basket than a lot of people were betting on. And this is the one or one of a few that that makes a lot of sense.

SPEAKER_01 37:49

Yeah, you know, I I was just looking at Liberty Energy. That was another behind-the-meter play we've done. Um, it's something I haven't been focusing on as much recently, but you know, that one's down. Um, it I just do think this space, like you said, is is going to be so attractive. It's worth mentioning too, Austin. One one thing I didn't talk about in this prior segment. We have earnings developing as as we're recording right now. Um, Core Weave report last night, Nebbius this morning. Nebbius is up 14%, Core Weaves up 22%. They um Nebius had mentioned they're signing large deals at 20 to 25 billion per gigawatt. Um, those are long-term deals. Uh, so it is interesting. This is just more kind of verification of this moment of what compute is going for, and it's driving those stocks up. Now, I I would say too, the other side of this all is if if we did have kind of a retrenchment in the market, these companies that are smaller have more tenuous financing um kind of profiles, it would be much harder on them versus an Amazon, uh, a Google, Microsoft. They're going to be able to weather that storm and they're probably going to be able to get out a reasonable amount of their deals and uh kind of rebalance. So, you know, it's just it's raising the stakes for everything. It's it's just higher growth across the board and it's it's a higher risk of uh slowdown happening faster if the spot market, you know, reprices.

SPEAKER_00 39:22

Uh so obviously one of the many macro items to watch is you said, you know, we we want to pay attention to the trends and then make investments in those trends. That's how you have performed so well. But now this industry is getting so big, this market's so big that we also have to start paying closer attention to some of the macro signals, you know, interest rates, spot prices, debt levels. I wanted to move on to the memory versus optics battle. Uh, financial Twitter uh had its, you know, the gloves off all weekend, and you know, the the bulls and bears were having it out. But memory and optics were two of the most salient bottlenecks over the last 18 months, two of the best performing sectors in your portfolio. So there's a lot of people maybe talking their own book because they made a lot of money on one of these trades. There's also a lot of people who probably have some sour grapes because they missed out on one or the other. So, what is going on with this memory versus optics battle? And where do you where do you draw your battle lines, Eric? Right? Which which which which which team are you are you choosing, or is there one?

SPEAKER_01 40:18

Yeah, you know, in general, I I don't post a lot on Twitter. You know, I we do we do put our profile on and sometimes we'll ask for questions and other things there. So if you haven't yet, you should go down and and give us a follow on Twitter because we will sometimes, you know, that is a place for engagement, it's a place to reach out to me directly. Um, but you know, financial Twitter, uh, I I know the general thought on Twitter is it's a cesspool. I completely understand why people would have that feeling, but certain segments of it continue to have just um aggregations of experts. And, you know, you can get both things on Twitter. You can have a feed that's entirely composed of people saying buy the stock and it'll be up 10x within three months. And you can have a lot of very thoughtful people with not many followers, but posting excellent ideas that, you know, sometimes only have 500 people following them. And and across the board this weekend, the debate was around this optics versus memory moment. It all started from um one specific personality, which is Jukin, who's an analyst for Citrini, and they're just widely followed, especially around memory. And they said that they thought that um optics was gonna have its moment, and memory is going to be in a position where it's it's going to struggle um at least temporarily. And their three main points on this were one, that memory is still seeing a deleveraging from Korea, uh, many people being margin cold and money coming out of leverage ETFs attached to memory. Also, uh LPs of some funds beyond situational awareness, I'm sorry, continue to be in redemption. Uh, they they so this would cause more sales than buying demand. Second, NVIDIA's reducing the amount of memory they have on Ruben systems, and they're going to focus more on tying these racks together with optics. So it would be memory's loss is optics gains. And third, they had referenced just a broad consensus. Memory will peak in the next two years. So, you know, this led to Austin. I logged in on Sunday and every single post was about this memory versus optics debate. And on Monday morning, it did flow into the market. When the market opened on Monday, um, the optics stocks were all up four or five percent, memory stocks were down four or five percent. And then within 15 minutes, the situation had completely flipped, where optics stocks coherent closed on Monday down 14%, and most of the memory stocks were up. So, you know, Austin, I would also say that's a little bit of a warning not to try and trade on these narratives you see online. Um, it certainly seems like some people were waiting to pounce on people buying into these areas and and taking an opposite position as you have people without a lot of conviction, maybe getting shaken out. But, you know, Austin, do I do I agree with this? Um, you know, first of all, this this dynamics coming about. I don't know if we talked about sand disk earnings recently, but sandis on their earnings, they said that they're effectively going to hold gross margins when they're where they're at. Um historically, gross margin peaks have been the sign to get out of memory. And if they're peaking, well, you're already normally too late, right? Because um the the markets already began shifting. And the question is are there new dynamics at this moment that could make memory today different than it was in uh 2022 and and 2000, all the different cycles it's been through? The first thing is, of course, The memory companies are signing contracts, long-term agreements. Well, Austin, that's that's going to cap gross margins, and we're seeing them come into effect. So that's trading off upside, but for a degree of certainty.

SPEAKER_00 44:15

Well, so also Sandisk's gross margins are what, 85% at this point? Correct. Yeah, a little below, but right around there. Yeah. You can't go much higher than that, right? Like just like so the argument that growth that gross margins are moderating is sort of like, well, like there's there's you you can't get more than 100%, and you're pretty darn close to that already. So I I don't give a lot of credit to that. I mean, yes, gross margins are inflated for all of the reasons we've mentioned, but to take the point that gross margins are l are are maybe plateauing a little bit as a sign that the industry is saturated and suddenly memory is not going to be a trade anymore or an investment for five or 10 years just doesn't make sense to me because they're at like supernatural super capitalism levels and they're holding there. And even if you see right, they are. Even if you see like some some erosion there, if it's if if it's oh boo hoo, like we're at 75% or 80% margins, that's still incredible.

SPEAKER_01 45:12

Yeah, and we've talked about this on the podcast. The historical peak previously was 60%. We were barreling towards 90%. So is it a much healthier environment in many ways for memory itself and for the broader AI infrastructure trade? If it lands somewhere closer to 75, uh, I think you could make a pretty strong case for that. Right. And Austin, if if memory prices did implode, there's probably two reasons this is happening. The the first would be some kind of technological breakthrough. Right, right.

SPEAKER_00 45:44

Reducing the need of memory for inference, um, which of course people are trying to do. But let's look at this NVIDIA reducing HBM on Ruben point, because I want to talk about that, because that that's something that the headline is really scary there. But my understanding is they're reducing HBM on Ruben, not necessarily by choice, right? It's that they they would like to put more memory in. They just, there's not enough supply, so they're forced to work around it. Is that correct? This is not correct. This is not something they want to do. This is just what the market is forcing them to do. And if memory pricing were to pull back a little bit and we're looking at maybe those 70 or 75% gross margins, which are enormous still, and more supply, it would be totally natural to expect that that NVIDIA starts to add more HBM back, right? This is this headline looks scary, but it's it's a result of the market forcing them to do that. This is not something they necessarily want. They want more memory, right?

SPEAKER_01 46:34

Correct. So, yes. So if memory pricing were to go down, if you had margins, you bring on incremental demand from things like NVIDIA using more of it. Also, consumer electronics, we're we're probably going to be approaching uh pretty extremely dear in smartphones and other things. If memory pricing came down, it would allow probably more growth to come out from there, some release valves. So I I think, yeah, it if there, if there was a technological breakthrough, um that could just be something that would impact memory itself and the broader kind of AI infrastructure trade could survive that. But Austin, there's nothing on the horizon, and there's nothing that can be manufactured at scale in any reasonable horizon. The second side would be if memory implodes, um, the pricing implodes. Well, Austin, that probably means that there are significant demand imbalance problems. And that means probably everything about AI infrastructure is in trouble. And you're gonna get just as hit as hard.

SPEAKER_00 47:37

You're at global economy issues there, right? At that point. At that point, we're starting to talk about China invading Taiwan or GDP slipping or recessions. Like at that level, there's much bigger problems happening, right? And and there's no safe harbor, is my point.

SPEAKER_01 47:53

Yeah, and let's we'll we'll talk a little bit more about optics. We've got Lumentup's earnings. I'm looking at them right now, they're up 14% following earnings. Though the one thing I would say too is um the point he made about NVIDIA shifting content from memory to optics um for Ruben does have some clear winners. And one of the recent stocks I recommended two episodes ago was SemTech Ticker SMTC. Um, and and that is a company that's going to benefit from this happening. So, you know, if if you are concerned to some degree about a company like NVIDIA making near-term choices to favor optics versus memory, well, we are making recommendations right now that tie directly into that. Um, so Austin, I I think it's just in my on this current debate of the opinion that you need to sell memory and and and get into optics. No, I think you should own both of them, right? And I think they're both advantaged. And I think memory, it has run up a lot. It is a trade that has stalled in many ways. It's similar to how people probably feel investing in NVIDIA right now. But I think NVIDIA is a fantastic. You know, I think if someone wants to just park money right now, if you have an excess amount of money and you're buying something, I think NVIDIA is a fantastic place to park that money, right? And it's not because you're playing a two-month catalyst, it's not because you're playing a three-month catalyst, it's because you think that across the next 18 months, the advantages NVIDIA has are really going to come to bear and you're gonna get a really above average return, but you're not going to get the thousand percent return. And and I I think that's part of it is I I saw someone saying on Twitter the other day, well, no one in this memory debate actually bought these stocks. No one actually bought Micron for $100 or less. Well, we actually did, right? So, you know, we we did own these companies from that point. Um, and and I think a lot of this is people have unrealistic expectations for companies. If if you were buying memory today, hoping for a 25% return into next year, that's a fantastic return for an individual stock across the past year. But when a lot of the people who have crowded into it were doing so on the expectations of multiple hundred percent returning the past year, well, that's not the era we're in anymore. That's not the era for NVIDIA. That's not the era for memory. Um, and and you know, if if that's the expectations people are going to set, they're going to be massively disappointed.

SPEAKER_00 50:26

Let's let's continue talking about NVIDIA because I want to hear about this half a trillion dollar deal. Because that feels like one, you know, the company is just so big, so further growth is harder to come by, which is why we're not getting those multi-hundred percent returns. But there's also a lot of overhang, I believe, in NVIDIA stock, in how entangled they are in financing. And that that's that's not necessarily a good or a bad thing. In many ways, they are backstopping this entire industry. But in the same way, maybe a little over a year ago, I made a comment that this, the entire AI industry felt like it was propped up by Sam Altman's deal making. Uh, fortunately, that you know, the the pillars diversified a little bit. We've got anthropic and TPUs and many, many other supports to the AI industry now, which is great. But the biggest pillar in the AI industry right now is NVIDIA's financing, enabling a lot of these build-outs. So there's a half a trillion dollar deal that could be overhanging the stock because people are gonna say, Oh, this company's so big, they're they're backstopping and um seller financing all these deals. So that can also be one of the reasons that shares are in having a soft moment now. But explain to me what's actually going on with this financing and is it something to be worried about, or is it actually an opportunity for investors like us to pick up a world-class company while everyone else is worried about the implications here?

SPEAKER_01 51:48

Yeah, it is funny too. I just pulled up NVIDIA's chart while you were finishing there. And it is as of this moment, it's it's up 18.9% uh for the year. We're we finished seven months out of the year that would annualize to a 30% return. I mean, any $5 trillion company, a 30% return. And it is, I think, just because the backdrop, of course, is so bullish. Semiconductors as a group has outperformed so much more. But in terms of what for many people would be an outsized holding, especially it is for me, it's my number one position. You know, you'd sign up any day for a 30% return in a given year. But Austin, yes, a $500 billion um investment. Uh, I don't know why it would be, I guess an investment partnership fund. I don't know what the exact they just announced this yesterday. There's so much to untangle from it, but essentially they partnered with every single who's who on Wall Street of private equity, etc. And the idea is they're going to turn compute into an asset class. Uh Austin, so two things on this. Number one, I saw an interview with Jensen, uh Jensen Wong and and all of the executives from Wall Street on CNBC the other day. I think it was the CEO of Goldman Sachs. He was like, This is this is the very beginning, like what was when I started mortgage-backed securities in the 1970s. And I understand his point. Um, that's that's probably also gonna raise people's hackles. You know, thinking back to poor word choice. Poor word choice, poor word choice. But you know, what this accomplishes is a couple things. Um, it it's going to create SPVs that you're going to be able to invest and compute like an asset class. Um, second, the fact that's an SPV, it's gonna largely stay off NVIDIA's balance sheet. In in terms of, we talked about last week, a lot of these numbers get tossed out when NVIDIA's doing deals and partnership, and people take the absolute highest number possible and anchor on that. For this one, they're providing a 25% backstop as part of this. So, you know, Austin, what would be their you know, capital at risk here? It would be 125 billion. Um, if things were kind of detangling very quickly in this compute market, they would have much bigger problems in the fact their revenue would be imploding at the same time. So I I don't think this is necessarily something that is going to be an existential risk in NVIDIA. And what I think they are doing, you know, this is we've we've talked about so many times recently this ceiling of what the AI market can reasonably spend in a given year. I've talked about, you know, Gavin Baker talked about a baseline of $1.3 trillion. And I talked about my estimates for $1.5 and it going up to $2 trillion. I said, the ways that this ceiling grows and we're going to be able to see growth rates beyond 2027 into 2020, 2029, 2020, 2030, is you're going to need revenue growth from AI itself, which we talked about, the kind of revenue growth that companies are seeing from inferencing at the current amount of um compute we have available earlier in the episode. The second area is things like this, just new forms of financing that are going to make capital available to these companies. And Austin, what I think the goal here for NVIDIA, this isn't money that's going to be utilized by Amazon. This isn't money that's going to be utilized by Google. This is really for the neo cloud type companies. This is giving them a financing advantage to continue going out and building and competing with these large companies. So Austin, I think it's it's interesting. We started the week, and for the neo clouds, um Amazon's or sorry, uh, SpaceX's, we're going to build as fast as we can, 300 to 500 billion. That caused a lot of these neo clouds to sell off because now they have a competitor with that ambitious of a target that's aiming to move that fast, which is, you know, putting their slice of the pie at risk. Then we see this in NVIDIA news, which is good for them. They're going to have access to finance. And now we get to the tail end of the week where we have Nebbius and Core Weave reporting earnings. And with Nebbius reporting that they are actually getting significantly higher rates for their compute, we see these stocks taking off. So it's been a bit of a roller coaster. And as I mentioned earlier, the near-term trends for these companies is definitely all up. Um, they they they they get this NVIDIA news, which is very good for them. They they get the increasing price of uh compute. And if if you are buying these companies, it is just a question of maybe what kind of risk is maybe being built into out years for them right now. Um, they are just entirely at this point a play for better or worse on what the what the cost of compute's going to be in the coming years. But uh for NVIDIA, this this move makes sense to me. You know, right now they they have half their business tied to these hyperscalers. They want that to be less, and this is one way of continuing to grow this industry across other players. So it it it strikes me as a relatively smart move. There's a lot to unpack, but I think this is this is NVIDIA, you know, it's a chessboard right now, and this is them moving some pieces.

SPEAKER_00 57:20

Yeah. Uh I I'd love to go back to something we talked about earlier in the episode, uh, or tie two points together, which is uh access to capital and the Taiwan semi, you know, uh suppressing force, you know, the rate limiter on this entire industry, of which you know we listed a few. But the counterbalance to that, the the counter narrative could be Intel, right? Intel coming online and manufacturing more. And Intel recently announced their first stock offering. So yeah, tell me about this moment here. This is another source of capital raising that is not, we haven't seen as much of it, but maybe we'll see a lot more going forward. So, what are they trying to do with this capital and what does this potentially mean for the industry and future capacity, appreciating that Intel, any capacity that comes online from Intel will take years.

SPEAKER_01 58:10

Yeah, Austin, they they announced that they're gonna do a secondary stock sale of $15 billion, largely going for CapEx. The stock was down originally on the news, but this was massively oversubscribed. I think they had more than $100 billion in interest, so it kind of shows us as much as people are asking right now, where's incremental capital coming from? Uh, a lot of the time when capital is made available, it has ravenous demand for it. At the end of the day, Austin, I I've talked about this continues there. They asked for $15 billion, they're gonna get $20 billion. Um, they they need this for their build-out. Um, they're they're getting this capital at pretty attractive rates, and the fact that there's more demand shows that if they needed to raise more in the future, it will be available for them. The through line for this is the companies in the semiconductor equipment space. And this past week, uh, we've got portfolio recommendations like Camtech and Anto. CamTech and Anto both report earnings. Camtech's up 19% across the past week. Anto's 28%. Both are kind of bouncing off this post-Leopold bottom. And the story for them is I I've talked about, you know, the vector that's going to matter for the future of building these factories that Intel's getting the money for, that SpaceX is going to build Terrafab for, that uh Taiwan semiconductor continues increasing the size, is going to be how we get kind of these chips closer together, how we package these chips. And this is a space that these are two companies that in their earnings said that they're seeing incredible backlog growth here. So, Austin, I think, you know, for Intel, nothing to be worried about. They had to raise capital. This is, they got it at pretty attractive rates, and this doesn't change the story at all. What it does change the story more for and continues to make look better is some of these semiconductor equipment companies that we've been talking a lot about in recent episodes, how good the tailwinds at their back are. And as I look at the earnings across this earnings season, it's generally very positive. And I think people are really catching on to how big the story is going to be the next couple of years.

SPEAKER_00 1:00:18

Wonderful. And yeah, I'm I'm certainly not a chip manufacturing expert. And chip manufacturing is the industry where the next big breakthrough has always been right around the corner and will fundamentally change everything. Some of those end up to be head fakes, some of them end up being true. But I mean, correct me if I'm wrong, one of the big promises that Intel has been touting for years has been backside power and it's very challenging to manufacture. But I think, at least based on my most recent understanding, they were sort of the sort of leading proponent of it and maybe had an edge on Taiwan semi there. So if that ends up being as good as they say it could be, maybe it maybe it changes the game a little bit in their favor. But let's let's move forward uh to Lumentum. I want to hear about your notes from the call. We talked about the optics versus memory trade. So give me your notes from Lumentum. What did you see in the call?

SPEAKER_01 1:01:05

Yeah, Lumentum's continues to be one of our best performing recommendations. Uh the we recommended it down, I think a little above $100 a share somewhere in that range. It's now trading 932 as I'm looking midday on Wednesday. Also up 14% today. Austin, when the earnings came out, it was down 5%, something like that. It's it's hard for these companies that are often really supply constrained to beat some of the whisper numbers, especially when they're training at the valuations they are. But when they hosted their conference call, that really shifted the narrative because what they had to say was so incredibly bullish for the future of optics. One quote I just want to highlight that they had on their call. To put this in perspective, for one major hyperscaler, the network capacity connecting just two AI data center sites could double the total global backbone capacity they built over the entirety of the last decade. So, Austin, sometimes, you know, we're talking about this putting SpaceX going for this 10 gigawatts, and it's easy to hear that it's hard to think about the enormity of what these companies are truly going for. That, you know, sometimes two data center sites being the capacity for an entire, you know, for entire industries in some cases over the past decade. It really shows, you know, you look at a stock chart of these companies and say, well, is this a bubble? It was flat for all these years and now it's straight up. And it's like, well, these industries often were flat for all these years, and now they have such extreme exponential demand that, you know, in some cases they've just got much more strategically important, especially in the case of optics. So, Austin, some things I liked about Lumentum's call, there's been lots of debate around future technologies in optics, such as co-package optics. We did a podcast probably back in May where there's a lot of reports about co-package optics being dramatically delayed, 800V dramatically delayed. Well, Lumentum emphatically rebutted this. They said that their visibility has actually sharpened for this market and that their largest customer, which is NVIDIA demand, is coming in the second half of 2027. We also saw NVIDIA putting out releases that they're going to have 800V products shipping in the back half of 2026. So we saw all this supply chain chatter. And I always say, don't focus too much on this because if you've gone through past launches, you see how messy this is. But if if the need for that product has a solid technological basis, it's likely going to happen in a form that's going to benefit the companies that we've picked for these trends. And that's exactly what we're seeing for Lumentum right now. You know, Austin, the competing technology for co-package optics is something called near package optics. Well, they said it's completely additive for us, significantly increasing their total addressable market. And, you know, Austin, speaking about, there's always a question about the competition from China. Um, they threw some pretty cold water on this. Applied Optoelectronics recently reported, and they had said basically what these companies claim to be capable of, off on their earnings calls or or in the media, doesn't match up with what the reality is. So Austin, the the other big thing is too, we look at the initial supply for a company like Lumentum, and on their call, they basically said that AXT, which is one of their suppliers, um, they they had to go to them uh recently to secure more supply, and they're going to need to do that in the next couple quarters. And they said that, and I quote, they are running well ahead on almost every metric on revenue, on margin, on operating margin, and they're going to give a new financial target at an upcoming conference. So I don't know what else is there to say aside from this is just really positive across the board. Uh, we we continue to hold our optics sleeve of not just Lumentum, but Coherent and Fabronet. And it certainly looks like the coming years, we talked about the structural reasons why optics had to continue growing as a percent of this data center pie. And everything looks to be on track. And a company like Lumentum has significant technology leadership in exactly the areas you want. So uh just a really impressive report. I think the market's reaction is reading this correctly. Wonderful.

SPEAKER_00 1:05:48

Eric, I wanted one more update from you, and that is on Microsoft's big ASICs bet. So it looks like they want to be working on some accelerators. They've worked with Marvell as the main design partner. So tell me about this news item here. And a couple of things come to mind as I as I see this story, which is Marvell was previously in the doghouse, right? Now at one point you had almost considered selling it, right? But we had always said, hey, you know, whatever your holding period is and investing, extend it. It'll serve you well. Marvell's an example where that seems to have played out. And it reminds me that the industry can trump operations. So Marvell was always perfectly positioned for the AI trade. And it took them a while because of their own operational shortcomings to catch up to it. But when there's that much pressure for the industry, right, the tsunami of the industry can overcome their own operational shortcomings. Um and it seems like Marvell is now catching up, right? It's become a great investment. They're at Microsoft is now working with them on these ASICs designs. So I'd love to hear about what is the news here about uh Microsoft's big custom chip bet. And do you draw any lessons about Microsoft, Marvell, or the industry overall from this news?

SPEAKER_01 1:06:57

Well, this is just an example of putting together various things. I know this is probably a little bit of a longer podcast. We've there's a lot of news this week, though.

SPEAKER_00 1:07:07

So whatever we buried, we buried a free stock pick. We buried a stock pick.

SPEAKER_01 1:07:12

Yeah, you're just hearing at the end because you skipped ahead. Let's put right here, let's put the stock pick right here in the show notes. And then when people come right here, they'll have to go back and find it early on. I'm gonna mess with the I'm gonna mess with the transcript.

SPEAKER_00 1:07:26

Let me mess with the transcript a little bit right now. There you have it. The stock that Eric is putting $200,000 into his biggest investment yet in the portfolio is Walmart.

SPEAKER_01 1:07:40

Um, but yes, so we talked about Microsoft that they're gonna have this AirPocket coming up, but they truly have incentives to scale as hard as possible that the market is not fully appreciating. One area that they're designing is their um custom accelerator, Maya 300. This is uh an accelerator that Marvell is a key partner on, and it looks like they're going to push Taiwan semiconductor. We'll see if they can get the uh, you know, the amount of capacity for it, but 300,000 chips in 2027, which is an absolutely massive number, Austin. And it it just does come back to if you understand Microsoft's incentive, Marvell being a partner, you're gonna be able to put together some of these tailwinds they have. They're also getting some strong tailwinds right now from the optics space. Uh, they had an acquisition of a company named Celestial that's apparent, you know, we're not far past it, but it's looking like one of the better acquisitions that's happened in recent years. And Austin, we're we're gonna talk about, I believe, in the next episode. Um, but CXL memory, they they have some massive drivers in a new market there. So this is a company that's that's just kind of got everything working for it right now. And I think we've got one more bit of news today, and then we can we can close this up at at uh just barely beating the uh 75-minute mark.

SPEAKER_00 1:09:01

A record-breaking episode, and our our listeners don't know this, but when we started the episode, I was like, gosh, Eric, you know, I'm feeling really under the weather. I got I got knocked in the face of a bug last night, and Eric is just taking, he is using that to drag me across the coals here of a record-breaking episode. But what is the last piece of news you wanted to you wanted to hit?

SPEAKER_01 1:09:21

Yeah, and for anyone out there, if you have survived this long uh on YouTube, a couple comments, some some cheers for Austin. He is he is quite sick right now. He he trooped it out so that we could get uh an episode filmed right now. So I I really appreciate that. And and a little love for Austin if we could, but Austin, last one, this is in your zone, so maybe you perk up a little bit, you get that second win. But uh Unit, which is by most accounts, you a leader in humanoid robotics. I mean, if anyone's watched the video on Unitry's various products, they're just cool. That's all these things are amazing.

SPEAKER_00 1:09:58

These things are amazing. I have lost hours on YouTube due to the shorts on these. These are the uh these are the Chinese-based robots um that perform. They're like the new version of Boston Dynamics, in that Boston Dynamics used to put out these videos that were fun to watch, and now Unitree's taking it up as a level, and they've got dancing robots and robots playing soccer and doing all sorts of fun stuff. But please keep going.

SPEAKER_01 1:10:18

Yeah, and they're they're IPOing. Um, there's some important caveats here. They're IPO. It's it's only $900 million, which uh when SpaceX is raising what was it, $85 billion? $85 billion. But Austin, it's 8,000 times oversubscribed. The amount of investor interest in this, and I don't even know. I I've seen some reports, 5,500, 8,000. I guess you could do the math, it's from a filing, but point being, it is so oversubscribed. I think, you know, recent major IPOs in the US, I think Snowflake was really notable. It was like 120 times, and Facebook was 20 times, and you know, those are some wow figures. We're talking thousands, thousands times oversubscribed. Um, you know, it's coming into a Chinese market that has been on somewhat of a furious rally. Uh, I haven't looked up in recent weeks, but I think that still holds true. Uh, a lot of retail fervor. So I don't know if you can necessarily draw a straight line to that, but it just continues to kind of put a focus on this robotics trend and um, you know, some of the interest we're going to see as it continues uh gaining traction across the coming years. So, you know, I I think uh an IPO that's 8,000 times oversubscribed says if if you're not really paying attention to the robotics market, you you really should be right now because um as as some of these end markets like humanoid robox begin achieving scale, there's there's gonna be a pretty furious um there's gonna be a pretty furious search among investors to find kind of who the leaders are across the supply chain for these stocks.

SPEAKER_00 1:12:02

Well, you you know I am the robotics guy, and I I do love this opportunity. And there there is a I do believe this is the next big great investment wave. This feels a little bit like self-driving in that it takes the leap from software speed to atoms speed in manufacturing, which is slower. So, in much the same way Elon Musk has been promising self-driving for you know uh 15 years and and being overly optimistic, eventually it did arrive, and eventually it was as you know as as good as advertised. It just took a while to get there. I think robotics is going to be similar. People have please.

SPEAKER_01 1:12:37

Oh, go ahead. Oh, I was just gonna say, as a small note, just on that, the recent recommendation we made in robotics, which was attempting to be a little bit more conservative.

SPEAKER_00 1:12:44

Oh no, uh Rego Rexford, yeah.

SPEAKER_01 1:12:46

Yeah, that that's actually down quite a bit. They had earnings that had disappointed a little bit. Um, so they're at 220 before earnings, they've dropped now down to 170. The the robotics they're facing two shifts. Number one, demand for modular data centers that they're extremely early on kind of achieving revenue from that. And also their revenue from robotics is is essentially non-existent today. And the question is how much it can scale. So, some disappointment into their earnings, but really unrelated to the opportunities we're interested in. So, I just do want to point out sometimes if this is a stock that you hadn't added to your portfolio. Um, nothing's changed in my opinion of it. It's it's being sold off for uh factors kind of beyond the core thesis of why we're interested in about a potential re-rating across the next two, three years. So that would be a stock I would look at at these prices and considering adding more of. So I just want to mention that in case anyone's hearing us talk about robotics and trying to associate with something that they can purchase right now.

SPEAKER_00 1:13:49

Yeah, and I think Ouster is giving a brief discount here, relatively. I think it's background 45. But I expect this industry will follow a very similar pattern as we saw with drones. So drones were very popular. I mean, at one point they had a uh sort of a hype cycle moment around even 2014. And then with the war in Ukraine and I think 2018, 2019, there was another sort of euphoric moment, and it traded roughly as an industry flat for three years, but it didn't mean that the underlying opportunity wasn't there. It just takes a while to manufacture these things. And there is a convergence of technologies that unlocks the capability there, right? There's a convergence of battery technology, of manufacturing, of uh the all the inputs into going into this physical hardware. So I think robotics are going to follow a similar trend where like the opportunity is app is as big as everyone has said it is, but it will be a bumpy and slightly longer road to get there because we're dealing with manufacturing timelines, which are longer and extended compared to software timelines, which we've now kind of been used to in this AI era where you know your thesis can play out in seven or eight months. That's not so with manufacturing, particularly with manufacturing that relies on tech on technology convergence, right? You need you need AI models, you need batteries, you need actual manufacturing, you need materials breakthrough. All of that doesn't mean the opportunity is not there, though. It just means that it will take a couple of years. And I would expect this is an industry that sees a lot of volatility as enthusiasm waxes and wanes. But just like the long-term demand for drones was up and to the right, you just had to be patient. I think the exact same thing is gonna happen here with robotics. AI-powered robotics, is my guess.

SPEAKER_01 1:15:31

Yeah, agreed with that, Austin. So should we let the people out? Should we dismiss class here? Let's do it. Let's do it. Record, record.

SPEAKER_00 1:15:40

Uh, thank you, Eric, for the marathon session. Thank you for our dear listeners who hung with us. We won't, we won't make you stay any longer. We would appreciate any comments, questions, or feedback that you can share on YouTube, Spotify, or wherever you get this. Uh, until then, uh, listeners, Eric, we'll see you next week. The AI Investor Podcast is for educational purposes only and should not be considered investment advice.