Threadguy Live
Stocks, crypto, politics, culture, and the great financialization of everything. Threadguy is live every weekday from New York with analysis, commentary, and interviews with leading figures across the space of internet markets.
Threadguy Live
Bubble Boi - Selling Everything, the Kimi Shock, and Killing DRAM
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Bubble Boi returns a quarter after his Intel and SanDisk calls went up only, and a week after his million-view "sold out of everything" tweet. He explains what one Chinese AI model changed about the entire memory trade, where his market exposure actually sits now, and the IPO he'd full port if he could get access. Plus why the worst memory maker keeps winning, whether China can really catch up on chips, the AI civil war brewing over open source, and the product he might build himself to kill DRAM.
And boom, Mr. Bubble Boy, the boy in Bubbles part two. How are you, man?
SPEAKER_02Yo, how are you?
SPEAKER_01I'm good, dude. It's good to see you again. Look, I want to start really quick. First of all, thank you for popping in on late notice. And second of all, I want to give you a couple flowers because I think it was last Intel earnings. So I guess about a quarter ago, you came on and you had one of our best performing stream guests with trade ideas. Intel, I believe Sandisk was the other one that proceeded to go basically up only since you first came on the stream. So congratulations on that. Thank you. Yeah, you're welcome. Um, well, look, thanks for coming on. We had a deal. If Intel beat earnings, you would come on, and Intel crushed earnings, but the market is in a little bit of a of a weird spot at the moment. And so I think it's actually more valuable for you to come on now. Well, things look like this. I mean, how how are you right now? Like, how is how is your year to date doing? How is your trading performance going in the market as we currently stand?
SPEAKER_02I think this is like the best year I've ever had trading or money-wise in my life. Whoa. Like I've never I think all the years I've been working combined do not add up to the money I've made this year.
SPEAKER_01All the years you've worked in your life combined do not make add up to this year.
SPEAKER_02Yeah.
SPEAKER_01Wow. Why do you think like this is just the year you finally put it all together and all of the work just caught up and it just hit? Like, why do you think that happened?
SPEAKER_02I I think it's definitely the AI trade is part of it. And it's just because of my like career and background, it was a lot easier for me to know like what was bullshit and what was not, and who was gonna make it and who wasn't. Um, because I don't like sit down every day like with the financials, and I'm like, oh, like this company's expensive, this company's cheap. I'm always just thinking about the products, yes, and the customers that are gonna use those products. So for me, it was just like right place, right time. And I've always been, you know, because I just like am an engineer, I've always been fascinated by AI and have been thinking about it for a long time and learning about it. So I was just in the right place. It's like all that research and knowledge that I've accumulated over the years is just like super valuable now.
SPEAKER_01I I know I asked you this last first of all, congratulations. I know I asked you this last stream. Are you still like are you a full-time trader right now or are you working a job?
SPEAKER_02Well, uh, I I I was uh probably consulting more back then, right? So I might have done like consulting gigs just because I like left my last job and I was just really wasn't into like going to a big corporation again or like some very like highly hierarchical place. So I was doing like consulting work, but I definitely don't do that now.
SPEAKER_01To your earlier point of you have this prior, you know, technical expertise that allowed you to see through the bullshit of what was real, what was fake, what was valuable at this initial blow off on the AI trade. Do you think that edge has eroded and there's sort of a new game being played? The tech technical expertise doesn't help as much, or is it equally as valuable as it was six months ago, a year ago?
SPEAKER_02It's definitely equally as valuable, and especially if you disagree with the market, right? So um it takes time for the market to either like understand new information, or the market might weigh like certain kinds of information more than you would because they're run by like people who are not as like in the weeds or technical. So um Intel Earnings is a good example always because no one really knows what to make sense of it for a few days. But I also think like an example of uh people weighing things more than they should is like whatever NVIDIA does is like all that matters, right? But people don't understand, like most of like the compute in the future, it's probably not gonna be NVIDIA, right? Like it'll probably be like uh all the hyperscalers' own personal compute, you know, TPUs, training, meta. It doesn't mean NVIDIA is like gonna go away. Like the pie is growing, right? Like the whole class of like market is growing, right? There's growth in this industry, but you have to understand that like what one person says doesn't set the you know rules for the whole entire industry, the industry moves based on what's the optimal choice, right? From an engineering or money perspective.
SPEAKER_01So, what do you think that means for the future of Nvidia's mode and like dominance?
SPEAKER_02I mean, when you're like when you have like these 75% margins to uh do like like right place, right time, right? But like as uh like Uncle Jeff always says, it's like your margin someone else's opportunity, right? Yeah, you can only get get away with it for so long, and there's a reason why hyperscalers for their own internal workloads are investing in you know their own ASICs, right? Like, or investing in their own interconnects, they want to own it because at the scale that they're spending money, why would you be beholden to a customer? Like, even you, like you want to like do a business, right? Like, you'd be really dumb if like you only prioritize one microphone company, right? For your podcast, you'd be like, Well, what if this microphone? Like, I like it, but I'm willing to change if like they start charging me too much or like they're not shipping. So I just don't I just don't know. I think um I think it's clear, like NVIDIA has you know a pretty dominant position right now, but the incentive structure for the labs and the hyperscalers is to you know vertically integrate. That's the incentive.
SPEAKER_01It's kind of like the Citrini post I think he made about uh with memory in Micron after their earnings. And he was like, okay, margins are insane, their earnings are disgusting, but the bear side to this is how long is the market gonna let them get away with I mean, fleecing everybody else. Of course.
SPEAKER_02And I've always I've always like, yeah, that's just to your point, like Micron has higher margins than even NVIDIA, and they sell a less differentiated product. Like people are more willing to um just think of the value NVIDIA is adding, right? Like they're more willing to pay a premium because you're managing, you know, infrastructure, you're making a unique differentiated product, there's all this capital expenditure. Like, people don't want to take on that risk all the time, but you know, that's why people pay a premium. But micron is really probably like the worst memory maker from just a quality perspective, really. Yeah, just like from a quality perspective, their HBM has always been like the lowest quality. And one thing, I don't know if I mentioned this in the last stream, but if you actually look at the big three memory makers, Samsung, SK, Micron, people think it's all about HBM because that's what the GPUs use. Now that's part of it, but most of the growth is actually coming from the commodity RAM. Because to make HBM, you take commodity DRAM wafers and you slice them, you dice them, and you stack them, right? So they consume commodity DRAM. So look at that. Um, you know, people still need DRAM for other things, right? But there's this big uh AI build-out where people are gonna just start consuming that. And the price of commodity DRAM has just skyrocketed even more than HBM. HBM is like steady to flat, maybe slightly up, slightly down, because they because they have capacity, right? But commodity DRAM, like just the stuff you're putting in your PC, that's shooting through the roof, and that's where Micron and all these other people are making their money. But really, Micron's the most egregious because their HBM was always the lowest quality. Their HPM is lowest quality, but they're just dominating on a layer below with the yeah, I mean like exactly because let's just say like NVIDIA, and I don't have these numbers off the top of my head, but let's just say NVIDIA sources 10 of their HBM from Micron, and they source like 40 from SK, you know, 50 from Samsung or whatever, right? Well, there's less DRAM, like there's just less commodity DRAM, and Micron's just gonna charge the market price for it. So, what does this mean for CXMT IPO?
SPEAKER_01Is that oh yeah?
SPEAKER_02All right, so like it's funny because you already because you said like, oh, I I made all these great calls. I'll give you a call. If you can get access to CS CXMT's IPO, you should just like full port that. Um, you should just like I was thinking of like, you know, pawning like everything I own to get more money to do this because it's extremely like mispriced, it's extremely lower. And the thing is, CXMT, despite them, like you know, they're Chinese, they don't really have an HBM product, the demand for just Kamai DRAM is so high, and they still get to charge that premium. And they're mainly selling to Chinese companies, right? So there's been like a lot of talk like CXMT is gonna crash the DRAM market. I mean, it's not in their incentive to do that either, because they don't make fucking like good HBM either, but they can sell to all the Chinese companies and make tons of money. And if you just look at like what micron trades at, what CXMT is gonna IPO at, and then just look at like how many bits or like DRAM sells their shipping, uh CXMT is extremely cheap, but that's partly because it's Chinese, right?
SPEAKER_01So at what price is it still egregiously cheap? Because I think hyperliquid valuation right now is 400 billion, give or take.
SPEAKER_02Yeah, that that doesn't sound that cheap to me, but I think uh the Chinese uh I feel it's 130, maybe. Yeah, that sounds pretty cheap. But I mean, even at 400, it'll grow into it. Like, I don't think it's the end of the world, but like at 130, I'm like moving to China if I can to get access.
SPEAKER_01Okay, I don't have that much experience trading emerging markets, but the the age old ism is like never trade Chinese companies ever. Why like why? Where does that come from?
SPEAKER_02I think it's I think partly it's because like government intervention, the markets like the government will back an industry, and then immediately when like the industry is in trouble, we'll just be like, you know, you guys gotta figure it out, like like stop living off of our like you know, welfare or our coattails. So the Chinese industry, like the uh you know, the CCP, the central uh party, they will prioritize an industry. They did this with uh EVs a couple years ago, right? And all the EV companies were great buys, but now they're not because they're in deadly competition. China's not about like, from my perspective, the Chinese like stock market and the government there is not like, oh, I want a big winner, I just want like the biggest market cap company. It's more like I want this industry and I want people to compete to lower the cost. I want like this is going to be a public good. Like electric cars, cheap electric cars is great for the Chinese people, it's great for the world. Cheap memory will be the same thing. But the thing is with these semiconductor companies, they actually can't employ that uh strategy. They actually have to back them because it's such like the capital that you need to play in this game, especially at the global scale, is just so high that it doesn't actually make sense to make people like compete. It makes sense for people to come together and work together, right? So I think uh for the semi companies, it's it's extremely like it's a great opportunity. And I'm invested personally in like some Chinese companies.
SPEAKER_01Okay, so speaking of investing in Chinese companies, I I might be misremembering this, but most positive it was you that were tweeting about buying Baba, and I have this like bit that's like every great every great man at some point in their life has tried the long baba and most and lost money, mostly lost money.
SPEAKER_02Probably every time I've long baba, I have lost money.
SPEAKER_01But they're in a crazy spot, right? Especially with this takeover of Chinese open source, Kimmy with Z Poo, like it feels like they are in dominant.
SPEAKER_02They're owners of all these like Chinese AI labs, they own everything, except maybe Deep Seek, but um I think also like if you if ignore that for a sec. If you just look at like Alibaba's like uh I would say like their e-commerce, and then if you look at their cloud, it's so cheap, it's so cheap, and it's still a very high-quality cloud. Like if you're outside the US, like if you're in Asia, if you're in like an emerging market, even in Europe, you might use like Alibaba's cloud. And now they have like the biggest, like basically the big one of some of the biggest AI companies in the world on there, and they're gonna start getting deals with like Huawei to get special chips. So I think like Alibaba's in a great position. If you think the Chinese AI market is gonna grow, like that that money's gonna flow through Alibaba servers.
SPEAKER_01There's uh Huawei is private, yeah.
SPEAKER_02Uh I'm not sure. I think they I think they have some public listing. I could be wrong.
SPEAKER_01Interesting on on uh on Baba. Okay, I want to go back to some of the broader market stuff a little bit. I I asked you your performance at the beginning of the stream. I don't want to make this like a like a uh a villain uh head to head, but Serenity posted monthly performance uh I guess like year to date, but in the month, it was on 49%, which is crazy because you I mean, yeah, like DRAM is down a lot, and Mags got hit pretty bad, and further out, bottleneck stuff side, all these things kind of got crushed, but like SPX is a couple percent, two percent, maybe three percent off ATH. Do you think that this is general performance right now? Like a mostly round tripped um PLs amongst the finter crowd.
SPEAKER_02So I hate to give like Serenity some credit here, but like uh you have to admit, like, most of these companies are down like 50% from their all-time highs, and it's like a broad industry sell-off. But I think people see that, and listen, like, we could all say, like, oh, like manage your rest better, do better, but like uh people see that and they may dunk, and I'm not saying like all of Serenity's picks are good picks, I'm just stating like these companies are up thousands of percent in this is great take. Let's say what's like even if they dropped 50 from an all-time high that you like picked randomly for a day, um they're still some of the best performing companies in the world, and they're probably will be the best performing companies for like the next uh at least for the next year or maybe longer, right? Depends on the company. But I think like when you see like sand disk or Intel moving 10 plus percent a day, like you can't be really shocked that you could go down like 20, 30, 50 percent. It's just the nature of the game for something that's high growth, right? Um, so that's what I would say about that. I don't think that's like a sign that you should just like sell everything, everyone's cooked. It's a sign that you should just probably like take some chips off the table and be careful. Um, so you don't actually go to zero. But you just need to like be uh prescient of what's going on. Like, think of dollar movements, right? Like Santa is probably probably about 50% down from its all-time high, but like uh in like actual like market cap terms, it's like barely moved versus where it was at the beginning of the year.
SPEAKER_01That's uh that's a good take. Let me see what it is. How much is actually down? I think it's like 40 37.4. Yeah, okay. So speaking of selling everything and panicking, yeah, I haven't screenshot it. July, July 17, 3 26 p.m. at Bubble Boy. I sold everything. Period. Enter, enter. Yes, enter, enter, everything, yes, dot dot dot, even that one. The next two weeks will be tough for most of you. Good luck. 850,000 views. Wow, which is a banger. I mean, dude, I I don't even think I realized how much um motion you got, which is 850,000 views. Yeah, congrats, first of all, for having that level of reach. But where are we right now with your exposure to the market? Like, is that still relevant?
SPEAKER_02Like, so net my portfolio, like I'm long and short, I have no positive exposure to like SPX. So, like on a delta basis, I'm negative SPX, right? Um, but not like largely negative, it's not like I'm shorting 100%, but I think it's like negative, like 15% exposure to the market. Well, and that's just because like you have to realize kind of like the game theory here. All these names are up hundreds of percent, some cases thousands of percent. Yeah, you have to understand like these institutional investors who bought them either bought them in the beginning of the year or halfway through, and they're like, if you bought Intel at the beginning of the year and you put 10% of your money in that, you probably made like a 30-40 return just doing nothing. That's just one small position. That is great performance for a professional money manager, right? That's some of their best years. Um, so they that like from their perspective, they need to lock in their gains and just like again, like you need to understand their goal is not to like become the richest person on earth, their goal is to get their bonus, like that's their goal, right? So it's kind of like positioning, but also it's like, you know, we we got like earnings came, the market ripped, everyone's realizing that the build out's real. Um, now there's questions of like, is CapEx sustainable from hyperscalers? Is CapEx by even some of these companies like overbuilt? A big reason why I tweeted that was like I for a long time was just like, you know, I had positive exposure, and like if I lost the days where like Sandis was down like 10%, I might have been down one percent on the portfolio max, right? Even though Sandis might have been one of my largest positions. Um the reason why I was very like move fast and just you know panic and sell is just because uh panic early. I think there was a lot of in uh innovations that came out, and it took me some time to really sit down and read through all the research of like what Kimmy's doing and all these other AI models, and also like I'm in the Bay Area, and not today, I'm in New York, but most of the time I'm in the Bay Area, and uh just like talking to people who are working on you know the research and just asking them questions of what like you know, do you think this technique's scalable? How long until this starts reaching other labs? And I started realizing like you have to like new facts have come. I cannot hold things with as much confidence uh with the prior thesis, right?
SPEAKER_01What do you I guess on the kim do the catbacks after on the Kimmy point? What do you think the emergence of Kimmy and you know the triple narrative violation? One, it's a Chinese open source company, two, the token cost is significantly cheaper. Three, they allegedly did it on Huawei chips. Like, what do you think the emergence of Kimmy has changed for your long horizon view of the AI trade?
SPEAKER_02So if Kimmy came out and it was just a really good model, but it's still using, like, you know, more or less the same kind of techniques or infrastructure of like the other AI models, I probably would have cut some risk, but I wouldn't have like really moved risk, right? I think it was that Kimmy's using a new technique that really cuts down memory growth entirely, so it's a fixed memory size, and they got really good results. So they didn't just like cut like you know, um, memory and got like okay results, they got world-class results. So the technique Kimmy's using for these, and we call them attention mechanisms and like tension mechanisms, yeah. Like these AI models have this thing that people have dubbed attention. Okay, it uses this technique called linear attention, and linear attention was used by the Quan models back in the day, or not even that long ago, right? But um, Quan wasn't like as competitive, right? So, like we didn't think much of it, and we didn't think like it would scale. Um, it's looking more and more likely that this scales pretty well, and if that's the case, you have a fixed memory size, it's always fixed, right? So the memory's not growing with the longer you talk to it, and the memory per user is gonna be fixed. Now it's not completely fixed, like a lot of pushback I got was like, well, that's only like 75% of the model is fixed, the rest still kind of grows. My point there was like, well, okay, now let's just say every model in the world goes to that. How much memory do we need? Right? There's a but since I've like looked at it, I've done more research, and I don't think it's as clear-cut, right? Like, it's it doesn't mean like we're done for, it just means like there's a trade-off that I think a lot of the Chinese labs and increasingly the American ones will start adopting.
SPEAKER_01Um, in theory, if every frontier model goes to this, the demand for DRAM drops significantly.
SPEAKER_02Well, DRAM for sure. I mean, HBM is different, but DRAM for sure, you just need to offload less so you can get away with it. Um, but here's the other thing it's like you can like everything fixed, because they already bought the memory, right? Like they already have these servers, everything fixed, they just serve more and more. So they'll make more and more money. It's not bad for them. They're going to make more money with the like imagine you had a car and like you just wake up and someone's like, Oh, I found a new way to drive it, and now it's like 10 times faster. It's like we'd be like, Well, why sell the car? Right? Like, yeah, it's great. I'm not gonna buy a new car though. Yeah, yeah. So I I thought it was kind of a bit of that, but also I just think like you have to really like readjust the assumption now that they're gonna like crazily spend for the next two years. What's to say there's not another innovation that shrinks this more, and now like you have this one uh server rack and you go from serving like a thousand users to a hundred thousand users. How can you justify spending 500k for a new server rack? It gets harder, right? Now, that being said, we have like two years before that that actually shows up, but it's uh it's definitely in everyone's conversations now. And I will say, even if this happens and they adopt Kimmy, the money will shift to different things. And it as a lot of people have already said this, and they've been right for the wrong reasons. Kimmy's actually really good for NVIDIA because the big thing for NVIDIA was most of the time when you're doing inference, the GPU is sitting there doing nothing. So you paid for this expensive GPU and you're just like not using it. Yeah, you're just not using it. Kimmy makes the GPU more busy, so it's more justified to like you know, buy the high-end NVIDIA stuff. Got it, but I think everyone's reasoning is just complete bullshit and lies. So that's another point.
SPEAKER_01Okay, that's to your earlier point on hyperscaler capex build out. We had Google earnings on Wednesday. We had Google earnings on Wednesday, they announced a substantial increase on CapEx guidance, and I have the chart up right now. Google's down now 20 from ATH, it's down $30 from the 350 level where uh Berkshire bought like Buffett bought $10 billion of their, I think $80 billion dilution. And I mean, Mags got destroyed yesterday. Yeah, and so what how like at what point are you or is the market gonna step in and buy Google, or do we just like hate this CapEx increase? Period.
SPEAKER_02Market the financial analysts hate CapEx, like it just lowers free cash flow, and everything's free cash flow to them. A company is only worth its free cash flow from now to infinity, right? So if you have no free cash flow for the next year or two, come and you had to buy the company, just hold it for two years. You don't, you're in theory, not gonna make any money. Like if you bought the whole company and just held it for two years, again, that's theory, right?
SPEAKER_01Yeah, yeah.
SPEAKER_02But like, but like you know, like the pyramid wasn't built in like a day, right? Um, it takes time to do the right things, right? And takes money and investment. And I think uh the market will punish them for capex just because there's like boomer fund managers who have like finance degrees, and that's all they know. They don't know that the pie is bigger on the other side, yeah. Right? Like, what was the capex of like Christopher Columbus finding North America? It's probably pretty high. Yeah, once we got here, it was like worth it, right? We're probably like we're that's the model I'm applying to this. So I think the market will punish them as long as they keep guiding to higher capex. Now, that being said, I don't think they will. I don't think they I think for this year it was very clear that they were gonna finish the guidance of like a little bit higher until next year. I think uh next year is when you're gonna start seeing um things slowly hint to less. You mentioned Google, but Microsoft is the one that got really punished, and it's because Microsoft's plan, like their their whole strategy of like we're gonna burn CapEx and make these AI data centers. Investors don't believe it because they're like, for what? Like, for what? What's your product?
SPEAKER_01Yeah, right, yeah, like like what are you doing? Yeah, what are we doing?
SPEAKER_02Yeah, for copilot, and then they're like, Oh, we're gonna rent this to people, and it's like, well, the big AI labs are doing their own data centers and they're kind of like diversifying. So, like, why should we like believe you? Now, there's probably like a kernel of truth that like people will just rent from whoever, but that's where the market will punish them. Google, for the longest time, didn't get punished because Gemini was pretty good. Um, they have like they use AI for all their products, even not Gemini, like it's just a common workload, right?
SPEAKER_01So I'd say Google search, like they have that everywhere.
SPEAKER_02Of course. So the thing is though, like they can only do it so much. Like eventually the market will just put its foot down and be like, Google, you can't just dilute me another 20 this year, like you can't just do another dilution. Google, you can't borrow a trillion dollars, like like chill out. But here's the thing as they're bringing this capex on, the money starts coming. The money, like it's like it's like time to recoup your cost, right? Like, if I uh I'm trying to think of a dumb example, like uh I'm opening a laundry map, like I have to invest in these like laundry machines and dryers. It might take me like a year or two to get the money, but that's gonna start happening with these data centers. Like eventually, the money is gonna start coming. And by my math, it's a lot of money and it's high margin, it's extremely high margin. Like, we're talking like north of 90% margin sometimes for serving these models.
SPEAKER_01How much of it is dependent on one Gemini being good or frontier level, which it is not right now, and they're maybe getting double punished for it. But then there's the second over like this gray cloud, I feel like, over the industry of cheap Chinese open source causing tokens to be erased to zero. Like, what if the Google payoff like from CapEx isn't as lucrative as it maybe it felt like it could be a year or two ago?
SPEAKER_02So I will say we'll address the Gemini point. To an extent, Gemini not being great is like not good. Like, if they have a better and better model, it completely justifies the capex. Yeah, and let me explain because it has to do with your second point. The Chinese models coming and making good models isn't as big of a deal, I think, for the CapEx. Here's the thing if a model like starts improving beyond the Chinese models, like we're we're valuing them based on the frontier capability, right? So if the next Gemini model can, like, you know, I don't know, uh try and think of a high value thing, like find a new drug, right? Like that can cure people, what's the value of that Gemini model? It's whatever the value of the drugs are. Imagine you can find tons of drugs. Imagine it can make a movie, imagine it can like do all this other stuff, and we're on track, like we will get there, it's gonna take years, we'll get there, but the Chinese models can't like unless the Chinese models start come like we diverge and the Chinese models just start making new capabilities that the American ones cannot even match. That's scary. But we're but it's not a big deal that they can code better, like coding's already done, like coding's gone. The next step is gonna be like doctors, lawyers, uh, airlines.
SPEAKER_01So, like if if Kimmy came out and made some novel pharmaceutical discovery, then it's okay, yeah.
SPEAKER_02Sell everything, yeah.
SPEAKER_01Yeah, okay.
SPEAKER_02Run.
SPEAKER_01Got it. But until that happens, we America's, you know. Okay, so then what is this? What do you think this means for? I know I'm jumping around, but uh economics for open AI and anthropic IPOs. Like, did Elon just cook everybody? He got out the door first, he got paid, and then now it's there's no time left for anthropic open AI.
SPEAKER_02So there'll be time for them. Um, Elon was smart for going first, right? And moving quickly because 112 right now on SpaceX, yeah. Speed is very important for getting money, and the more money you have, the more you can fight. I don't think the IPOs are cooked, but if we're in like a real prolonged bear market, they are. I don't see that. Like, despite my tweet of like selling everything, I want to buy back, but I just think there's a time to do it, and then the time's not right now. So I think um, like the anthropic IPO shouldn't be derailed, should still come at the end of this year, I think. Right. Um, the open AI one we have to see though, because they don't want to IPO, and then immediately people are like, Oh, anthropic's model is better. Let me just sell you. Like, the most common trade in the world is gonna be like long anthropic, short open AI as like a hedge, right? Um, but that's but again, like this is this uh removes a lot of nuance.
SPEAKER_01So, are you a re-entry on like price capitulation or on some news-based AI advancement event?
SPEAKER_02I think the price needs to go down a bit, but also I think we need some like more clear signs from the AI labs and the hyperscalers that like they're not worried, like they're good. Um and also I'm just doing my own research of like what does these like new innovations really mean, right? Yeah, like what are the impacts to all parts of the supply chain? So part of it's me doing my work uh being more confident because when I invest, I invest with like confidence. I'm not like uh like an indexer, right? Like I'm not trying to just buy like one percent of every company, I'm just trying to put money in the company I like, right? So it's a bit of a different investing mindset for me.
SPEAKER_01What are you raising the fund?
SPEAKER_02What fund?
SPEAKER_01The hypothetical hypothetical fund.
SPEAKER_02Oh, I don't know about this fund that I've heard uh people talk about.
SPEAKER_01Okay, the one you know, the one other thing we haven't talked about, I know you gotta go pretty soon. Let's just go five, ten minutes. No, no, I got time, I got time. Okay, okay. Um, but the one other topic we haven't really covered is Intel, which is funny. Um, so I guess to start, like if I remember correctly, when you came on three months ago, give or take, um, your Intel thesis was really based off like their packaging ability and the new note.
SPEAKER_00Yeah.
SPEAKER_01Where are you right now on Intel as a company, its future growth, your conviction in it?
SPEAKER_02Yeah, I mean, I think it's by the way, I think it's like super bullish. Like the earnings call was great, despite the sell-off, because they're guiding for CapEx. The pre- the reason it's selling off is people think they're gonna do more capex than they're talking about. They're very admint that they only need to do 20 billion. And by my math, I think 20 billion is more than enough. Also, they mentioned in the call they have prepayments from customers, so customers are putting up the money saying, I want capacity whenever you get it. And they mentioned, forget 18A, which is already in production in 18AP. They mentioned they're going all in now on their next generation 14A. That's because there's a customer. They've all they've said this many times, they would never boost CapEx until there's a guaranteed customer. I take them at their word for this. Like, CapEx is no joke, especially for this, because once you put this stuff up, you need flow all the time. It's not like an AI inference cluster where like it's just API, like you need new customers, right? All the time. You definitely think that they have long-term commitments there. Um, so I'm very uh bullish on Intel. I think, like, honestly, like the market's selling off a bit, the CapEx rumors, like it's just like people don't know what they don't know, right? Similar to me, like Kimmy came out. I didn't know what I didn't know. I have to like cut some risk here. I think right now is like the best time. I mean, maybe not like today, today, but like in the next in the next six months to a year, like this is like you're getting it even cheaper. And it's like, by my math, 14A has a real customer. You could just do the math of like, okay, 20 billion, how many like machines does that buy? How many wafers could that translate to? And what would they charge for them? And you just start getting ridiculous numbers, you start getting numbers that are like a hundred percent revenue growth, like within a year. Disgusting. So I just think uh, yeah, and and by the way, I've heard um, I've heard like my my name's been ringing in these uh pod shops now because it's become a common like trade to like bet against bubble into the Intel earnings, uh, apparently. But I think uh people the the market discounts things until they start seeing results in the earnings, right? When it happens, it happens all at once. Like when Intel beat last time and they said, like, oh, we're positive margin, we're gonna be higher growth because of CPUs and more positive margin for longer. It that's all they needed, and then the whole move happened instantly.
SPEAKER_01We have this bit, we have this bit in the stream never bet against Bubba Boy.
SPEAKER_02I I think it's okay to bet against me. I'm not always uh right, but if I'm like confident in something, just know like I wouldn't be overly confident if I didn't have like good research to back it up. I am very fallible. Like when I know I didn't do my research and something new happens, like Kimmy, I very quickly reassessed my plan.
SPEAKER_01I think as you should. Okay, here's another side of the market that I don't understand that well at all. But I saw this, um, I I saved it just in case. The semi-analysis made this post about SMIC and how they're beating Intel's uh 188 18A. Like, can you break down exactly what that is and like can China catch up on that side?
SPEAKER_02You're asking the right person, I think, for this. Um, so to your second part of your question, absolutely China can catch up. I'm not one of these people who just think like EUV is like this like godsand that like no one can replicate. I mean, like, we're not idiots, like you everyone like the laws of physics are the same here in here in the US and China, right? You can only like uh hide things for so long. The question is more when do they catch up, not if. And the comparison of like 18A and SMIC, I mean, it was very disingenuous, I think. Um, the comparison is like transistor density and all these other things and pitch. That's not like performance, like that's just like one metric. So you look at one metric and you're trying to say like what the whole performance of the node is. I could pull out a bunch of other metrics that will say Intel's better. The other point is like 18A was like early, like eight, like they picked out Intel. Why didn't they pick out Samsung? Which was even worse than Intel, right? So it seems pretty targeted in my view. Um, I don't think they're like good friends of mine, so I think that's why they're uh trying to pull this uh wool over people's eyes. But 18A is like okay, it was their first node to ramp up, and they just started moving their real capacity, like their own internal capacity from TSMC onto 18A. And 18A's benefit is more power, it's not really like higher, more transistor density, right? 18AP, I'd like them to do the same comparison, and 14A, I'd like them to do the same comparison. So if they're like a real reputable publication, they should just keep updating and do like a node head to head, right? It's it's completely fine if like someone wants to point out like flaws. I just think like you know, like the way you present the information already biases it, right? Like uh the media is biased just because of choosing what to report on, right?
SPEAKER_01How much impact do you think they have on the market if they were to go full-blown, like you know, uh anti-Intel crusade? Like how much intel uh influence do you think they have over the market?
SPEAKER_02I think in certain scenarios they have a lot, in certain scenarios they don't. I think if they went on a whole Intel crusade, it doesn't really matter. I mean, Intel's gonna spend the money and make the money, like they have cut like like no, yeah, at the end of the day, like like uh Google and like NVIDIA are not gonna read like the semi-analysis report and be like, yo, actually pull those, pull those ways from Intel. Like, it's not how it works, it's more like the investor community, right? Yeah, yeah, and even then, like everyone needs to do their own research. Like, if you're not doing that, what's the point? And I've heard from people who like are professional like hedge fund managers that they ban some of their employees from reading Southside research, they ban them, they just say don't read it at all. So, Southside Research is just there to color what's going on, it's not really there to like tell you what to do.
SPEAKER_01Yeah, I like this take. Um, okay, so there's this another thing that happened that I think was kind of fascinating is I guess it was a a month or so ago now, but the last Leopold Aschenbenner situation awareness 13F was uh I don't know, uh eye-opening, I suppose is the term that I'll use. And then it was later disclosed the Shaw's 20% investment, which I'm now a community member bag holder of at like, I don't know, one and a half billion dollars or something from Australian Neo Cloud.
SPEAKER_00Yeah, yeah.
SPEAKER_01I don't know. What do you think about the direction that Leopold and it it it sort of resembles like the AI bottleneck trade a little bit? It's like okay, fuck it, we're doing photonics. Okay, fuck it. What's more what's cheaper than photonics, but now it's just the money managers are doing it. What do you think about that?
SPEAKER_02So I have to preface Leo's a good friend of mine. I don't give him any advice or anything, but we've uh we're we're mutual respects for each other. I think I think he's not like uh he's more sophisticated than he comes off, I think. Um, his investment in neo clouds has a very, very specific thesis, and I actually don't think it's wrong. Depends on the neo cloud, but like he has a very good thesis on this, and also like if you're gonna invest in like flash memory, photonics, all these things, they're by definition cyclical, right? Like, you're gonna invest, you're gonna build out, and then you're gonna wait till you need to do it again. He's has a very specific thesis, and I don't really want to state it for the public because I'd rather him uh state that when he's ready. Like, I don't want to like give away his like story, but I I I I don't think it's like non-sound. I mean, I'm pretty bullish on certain neo clouds, and when I started uh moving my portfolio, one of the first things I bought was neo clouds. So I don't think neo clouds in the short term are gonna do bad. The question is just like long term, do these neo clouds just mature into real clouds, like real infrastructure, right? So there's a lot of uh debate, and there's a lot of smart people, I think, looking at it. Um, I wouldn't say like my whole portfolio is long neo clouds, though.
SPEAKER_01You should become a Shaw's shareholder with me. Yeah, I'll look into it. I didn't really, I just saw a situation awareness 19.9%. Okay. Um, another interesting thing happened today is um Jensen made a Twitter account and threw out a war cry rallying on the side of open source AI. Yeah. It feels like we're setting up for this artificial intelligence civil war that is basically Sam and Dario who hate each other, but maybe they have common enemies against maybe they have Trump Admin on their side, but they don't have David Sachs against the open source Chinese companies, which are quickly become like the topic of just the topic. Um, what do you think happens?
SPEAKER_02I hate to agree with Jensen because he is talking his book, right? Like, this is good for NVIDIA, right? So good for NVIDIA, and he loves neo clouds, he he wants neo clouds to exist, he doesn't want all the compute in like a handful of people's hands, and other people can't come up with models. If anything, he wants everyone to have local compute, every business. That's the benefit for him, right? I will say, like, yeah, there is gonna be like a bit of a civil, as you say, like a civil war. Like, a lot of people think like we should ban Chinese open source models, and they cite all these bullshit risks and reasons. I think um if you're an AI lab, you're a lap, like lab is in your name. Yeah, what are you doing? Like, your job is to innovate. If you can't like innovate, you're not gonna be a lap for long. So I think you should focus on what America's best at, and America's best at innovation. China's great at iteration. China's so good at taking something, making a thousand copies and refining, refining and making it better. Like, um, like one of my like favorite things to do in like my pat in like my free time is like look at YouTube Shorts, and there's all these YouTube shorts of like Chinese, ancient Chinese construction, and they like go like how they built their bridge, and like the first time it failed, and then like a thousand years later they got the perfect bridge. Like, this is like built into the Chinese culture, they're so good at refining things over generations. So, I think, like, yeah, like long story short, I just think the AI labs like trying to ban competitors is really stupid, and it's gonna happen with or without them. Like, we're not gonna have a monopoly or a duopoly in AI. This is not a thing. There, there's many reasons I could state. One of them is data, like you need certain kinds of data to do the next leap in ability. Like, how are you gonna discover drugs if you don't have data on biology and chemistry, right? Um, like you're not gonna do it, right? Like, you could you could only like do so much just like in your house reading textbooks with a pen and paper, right? So I think uh they need to realize that their job is like not to like fight the low-end consumers, their job is to move the needle. And if you can't move the needle, then you shouldn't be in your business in this business at all.
SPEAKER_01YouTube shorts, I hate to just pick this one out, but YouTube Shorts and the big two six is a little bit questionable.
SPEAKER_02Yeah. I'm addicted to them for for a weird reason.
SPEAKER_01You know, there's also this funny, like, there's this deep irony where it's like um the US wins the internet by Being open and letting everyone basically build the internet on top. And then round two, you kind of get like a startover, and the US is like edging towards like closed AI, or at least pushing for closed AI. And China is like, no, no, no, we're gonna do open AI, and like Xi does a speech, and he's like, No, no, no, we're doing open AI, like yeah, open AI. Um, it's yeah, but deeply ironic how that's playing out.
SPEAKER_02Yeah, I I will say though, I don't think like a lot of people have told me they're gonna ban Chinese models. I don't think there's any chance that's happening. I just like from the people in the administration who actually um advise like Trump, and of course, there's Congress and the Senate and all that, but there's just not really it doesn't really make sense.
SPEAKER_01I mean, what do you think of the trade if that happens? We've I've spent a lot of time trying to think of this. Wow, and I originally went to the case.
SPEAKER_02They can't really ban it anyway. Like, we will just fly to China and come back with some flash drives. Like, we'll figure something out. The question is more like if you work with the government, how can you um ensure that these models are safe? So if they ban it, they'll say you can't work with the government if you're just using their API. That's fine, that's that makes sense, that's reasonable. But a whole bunch of companies will come out who say, Hey, we downloaded Kimmy and we're gonna run it safely for just for you. Like, that's gonna be the new thing.
SPEAKER_01What is your thesis as to why there's no chance the US will ban it? Because I don't think that that's consensus idea.
SPEAKER_02I mean, um, you kind of see like Jensen made his whole post. Do you see the people who are signing on it?
SPEAKER_01Yeah, the biggest Elon posign, everybody.
SPEAKER_02Yeah, Elon, like Elon doesn't want like if you can beat Elon, go beat him, he'll try to get better. Like, that's the whole point. Like, this is a competition. Like, why are you trying to do unfair tactics? I think even Sam could it's really just Dario, right? Sam, uh yeah, of course, yeah. Like, even um, yeah, I don't think Sam hates it at all. I think it's more like, well, you know, like we're focused on doing the next big thing, we're not focused on just like commoditizing, like you know, like little uh Python scripts, like that's not gonna be where we make most of our money in the future, right?
SPEAKER_01Um okay, I'll give you like uh one final and then like a sign off. Um, even if you're not long anything crazy right now, what are the emerging trends and or trades that you think are the most interesting over the I don't know the rest of the year?
SPEAKER_02I I said I think I hope I said this on your last stream. I said photonics is cooked.
SPEAKER_01You did say photonics is cooked, yes.
SPEAKER_02I was absolutely right, and people keep trying to like uh grab for straws there. I will say though, there may be like a new way to do like quote unquote photonics that is less cooked, um, but it's gonna not be any of the names that people have right now with like these very expensive lasers and advanced like TSMC stuff, it'll be very different. So I think um that's uh something that I'm excited about, and I think NeoClouds are in a very interesting place right now. I myself am a customer of like Nebus, right? Because I want to run Kimmy and like figure it out. The only way I can really run Kimmy is if I get one of those big servers.
SPEAKER_01How much is it?
SPEAKER_02I don't know if I could say because it's like contract to contract, but we're talking like hundreds of thousands a month.
SPEAKER_01Well, what the fuck are you doing on it?
SPEAKER_02We're doing research.
SPEAKER_01Whoa, yeah. And what is the it's like, why do you have to run apps? We kind of glossed over that by the way. Yeah, um what what is the like it's break down the advantages of this approach running locally like this?
SPEAKER_02Well, for me, it's more like I need to like, I'm not gonna like take someone else's word of like what Kimmy's impact is gonna be. I'm gonna run it and I'm gonna say, okay, like how is it running? Like, what what do the tests say and all this? And the other advantage is just like you run it for yourself, you can do all kinds of things, you can change it, you can optimize it. And also, I think uh when Kimmy starts becoming really open source, like when they drop the weights, it's gonna be a fun uh race to see who can get it to run because it's not that easy.
SPEAKER_01So I would like we spent a lot of time talking about on the stream talking about local AI and what is the best expression to long local AI, and I'm always like fuck, you know. I wish um, you know, I wish like Mac Mini was its own company.
SPEAKER_02By the way, the answer is Flash for long AI for long local AI, it's flash, and who specifically? All of them, all of them, yeah. I think uh there's a time we need company your readers should look into called physon, physon uh that make flash controllers, and they're getting rid of DRAM. I'm very bullish on this idea of getting rid of DRAM. Who knows? I might be doing it myself.
SPEAKER_01Whoa, it can you give us like the highest best explanation you can as to why it's flash?
SPEAKER_02Well, flash has like density, right? Like, like you can store like one point like terabytes on flash, right? So there's that. Also, if you look at like what you're really doing during AI, like all the naive like EEs will tell you, like, oh, flash is slow. Well, it's it's slow if like you just use it for anything, but if you use it for AI, you need to know how to use it, how to talk to it. AI doesn't always need fast memory, it doesn't always need this. You have to know what you need and when. So the access pattern matters more.
SPEAKER_01Okay, flash, anything, any other themes that you really like or specific names coming here.
SPEAKER_02We should uh not this year, but next year. I'm very interested to see what Qualcomm does.
SPEAKER_01What is up with every company being like, oh, we're building some uh device with Qualcomm? There was a SpaceX League, then back, there's like the opening out.
SPEAKER_02Why what they have some sauce, man.
SPEAKER_01They just have it, like what is it?
SPEAKER_02Well, I I was very skeptical, and I think everyone's very skeptical of Qualcomm right now, so we have to see what they really ship, but there's reason to believe that they're not like completely inept and liars. Um, they may actually like if you like you could pull off pull up on your stream eventually, like Qualcomm's like data center accelerators, what bandwidth they're quoting, right? Okay, they're quoting ridiculous bandwidths, and many people just don't think they can hit it. Uh, they just don't think it's real. Like, they think like there's some like you know, little terms of service, little asterisk there. I think there's a chance that they might actually not be joking.
SPEAKER_01And if they're not joking, you think it's it just goes crazy.
SPEAKER_02I think it has a huge advantage, yeah.
SPEAKER_01Nice.
SPEAKER_02Balcom's a terribly run company, so let's why, why, why, why it's just this terrible management, like just really like they've basically alienated alienated like every customer they've had that like they alienated Apple, they got a huge lawsuit, they had a practical monopoly on like modems for like uh 4G and 5G, and now they need to like reinvent themselves. But they're like cockroaches, like they never die.
SPEAKER_01So that's hilarious. I didn't really have the lore on Qualcomm.
SPEAKER_02Like, yeah, that's Qualcomm and Apple had a lot of beef, and they they they sued each other a lot.
SPEAKER_01So who won? Do we know who won? Apple. Apple won, of course, right? Fuck. Of course, right? Um Boba, you're the goat. Okay, give me like a um I don't know. I guess you gave me a sign-off. Like, I don't know, what are you excited about? Give me this as a sign-off. How little does DR how low does DRAM go? And like when are you, I don't know, what are you working on or what are you excited about right now that you haven't mentioned, if anything?
SPEAKER_02I'm excited for Flash to take what DRAM's job is. So when you start seeing that happening, sell DRAM. Like, like we're not there yet, but yeah, I mean, like, you can only charge people extortion for so long.
SPEAKER_01So okay, actually, last question give me a um a sign off on uh favorite trader on Twitter right now that you think it's is hot or underrated or doing well or anything. I have to pull out my phone and see. Who's the next bubble boys? What I want to know if it's possible.
SPEAKER_02Sadly. I think um I think uh irrational analyst is a very smart person, and I think he's a real like photonics expert and an engineer. And I would definitely like don't just take him at it, like what he posts, like at his word, but like read what he says and read what he talks about, and you'll start learning about the industry, like in photonics. You'll probably make better investments. And who knows? Like, if like I said, like if photonics gets uncooked, it's gonna be this guy who finds it, right? He'll find you the stock, right? So he has some recommendations, but we're yet to see if they've played out.
SPEAKER_00Nobody's not waiting.
SPEAKER_02We like hang out all the time, like in the base. Yeah, me and him, we hang out all the time.
SPEAKER_01You got some chrome on right now?
SPEAKER_02No, no, no. You didn't bring any chrome? No, not today. I I'm in the city, it's kind of hot. I have to run around.
SPEAKER_01It is kind of hot. I don't know if Bubble was really bad. I've never seen him fucking hit a money spread in some chrome jeans with a class azul in the club. No, no, not yet. Not yet. Um, dude, you're the go. Uh part two was sick. I appreciate you coming on, man. Um, thank you.
SPEAKER_02You're a shout out because I tried to shout someone out last time, but I messed up shout out, my boys at through shout out through.
SPEAKER_01Maybe one day we'll do a thro a stream on what is through. Guys on, Robert Chang, yes, sir. Bubble Boy and Robert Chang, Titans of the cryptocurrency industry, is we're in good hands, man. Shout out to Thru, Bubble Boy, shout out to Bubble Boy. I look forward to part three at some point in the near future. Maybe $200 intel. And um, maybe I'll see you in New York. Yeah, for sure, man. All right, my guy. Much love, dude. Thanks for coming on. Bye. Peace.