1 00:00:00,020 --> 00:00:03,860 Ejaaz: Gavin Baker is one of the most prolific AI investors that almost no one has heard of. 2 00:00:03,960 --> 00:00:08,160 Ejaaz: He spent the last 20 years investing in some of the biggest AI companies before 3 00:00:08,160 --> 00:00:09,520 Ejaaz: they became household names. 4 00:00:09,660 --> 00:00:13,900 Ejaaz: He was an early backer of NVIDIA, as well as Cerebris, which IPO'd very recently. 5 00:00:14,020 --> 00:00:19,160 Ejaaz: And he has a very concrete thesis, which is AI isn't in a bubble. 6 00:00:19,360 --> 00:00:20,100 Ejaaz: It's quite the contrary. 7 00:00:20,320 --> 00:00:23,960 Ejaaz: It's in a super cycle. He says that by looking at the watts, 8 00:00:24,200 --> 00:00:26,960 Ejaaz: wafers and tokens, the infrastructure of AI. 9 00:00:27,160 --> 00:00:29,780 Ejaaz: He's identified some of the key bottlenecks and constraints, 10 00:00:29,780 --> 00:00:31,420 Ejaaz: and he has one simple thesis. 11 00:00:31,620 --> 00:00:35,360 Ejaaz: The biggest returns that you can get in AI is in electricity, 12 00:00:35,760 --> 00:00:37,460 Ejaaz: power, and silicon fabrication. 13 00:00:37,500 --> 00:00:40,380 Ejaaz: It's got nothing to do with SaaS software as a service. 14 00:00:40,600 --> 00:00:43,660 Ejaaz: It's got nothing to do with chatbots, such as Anthropic or OpenAI. 15 00:00:43,860 --> 00:00:45,960 Ejaaz: It all filters downstream from 16 00:00:45,960 --> 00:00:50,200 Ejaaz: semiconductors, the picks and shovels that build the entire AI industry. 17 00:00:50,320 --> 00:00:53,840 Ejaaz: And he's been expressing that interest to the tune of $4.1 billion. 18 00:00:54,200 --> 00:00:56,880 Ejaaz: While most people are calling the AI industry a complete bubble, 19 00:00:57,300 --> 00:01:01,760 Ejaaz: he thinks that it is a generational buying opportunity specifically for AI infrastructure, 20 00:01:01,760 --> 00:01:04,300 Ejaaz: and he makes the thesis very clearly in his fund. 21 00:01:04,840 --> 00:01:08,420 Josh: And if you hear these constraints that he's talking about, this AI infrastructure, 22 00:01:08,880 --> 00:01:11,500 Josh: you realize that this kind of sounds pretty familiar. 23 00:01:11,620 --> 00:01:14,140 Josh: We've heard this thesis before, and that's because on the show, 24 00:01:14,180 --> 00:01:17,860 Josh: many times we've covered an investor by the name of Leopold Aschenbrenner. 25 00:01:17,960 --> 00:01:21,120 Josh: He has been around for three years. He just published his most recent 13F, 26 00:01:21,240 --> 00:01:23,480 Josh: and he has a lot of the same core philosophies. 27 00:01:24,460 --> 00:01:27,720 Josh: Now, here's the difference. Leopold's been around for, what, three years? 28 00:01:28,000 --> 00:01:31,360 Josh: Gavin's been doing this for 20-plus years. And when we're comparing these two, 29 00:01:31,800 --> 00:01:34,220 Josh: Leopold actually has almost three times the assets under management. 30 00:01:34,400 --> 00:01:37,160 Josh: But there's this great quote that I was reminded of from our producer Luke before 31 00:01:37,160 --> 00:01:39,760 Josh: the show. It's like, okay, you can beat Warren Buffett over a year, 32 00:01:39,860 --> 00:01:41,740 Josh: but can you beat him over multiple decades? 33 00:01:42,000 --> 00:01:45,860 Josh: And Gavin Baker has a track record that proves that he might have a slightly 34 00:01:45,860 --> 00:01:48,200 Josh: different outlook on this investment thesis. 35 00:01:48,460 --> 00:01:52,200 Josh: For those who don't know, Gavin Baker, he is the basically founder of this company 36 00:01:52,200 --> 00:01:53,120 Josh: named a Trades Management. 37 00:01:53,340 --> 00:01:57,520 Josh: They are an investing fund, and he has spent 20 years investing in NVIDIA. 38 00:01:57,520 --> 00:02:01,720 Josh: Now, if you've invested in NVIDIA for 20 years, it's a miracle he's still working 39 00:02:01,720 --> 00:02:03,260 Josh: because those are some pretty incredible returns. 40 00:02:03,420 --> 00:02:07,240 Josh: Some of the early wins that we've recently seen include companies like Cerebris, 41 00:02:07,360 --> 00:02:08,880 Josh: which he was a very early investor in, 42 00:02:09,440 --> 00:02:13,580 Josh: Cerebus, if you'll remember, just IPO'd for an ungodly amount of money. 43 00:02:13,820 --> 00:02:16,960 Josh: Same with this company, Estera Labs. And there's a lot of other companies that 44 00:02:16,960 --> 00:02:20,480 Josh: I don't think you've probably heard of before that we're going to cover in this episode. 45 00:02:20,600 --> 00:02:25,160 Josh: As we walk through his portfolio and his guidance on where he sees the AI industry 46 00:02:25,160 --> 00:02:28,340 Josh: going in terms of investment and where all the opportunity lies. 47 00:02:28,560 --> 00:02:32,360 Ejaaz: So then the question becomes, what is he investing in and why is he investing in it? 48 00:02:32,460 --> 00:02:35,160 Ejaaz: Well, if we look at his recent 13F for a tradies management, 49 00:02:35,380 --> 00:02:38,680 Ejaaz: which is the name of his fund, They have about $4 billion worth of AUM. 50 00:02:38,800 --> 00:02:43,640 Ejaaz: We look at some of his biggest positions and unpack what some of these companies actually solve. 51 00:02:43,780 --> 00:02:47,800 Ejaaz: It kind of points towards the bottleneck that Gavin references in a bunch of 52 00:02:47,800 --> 00:02:50,460 Ejaaz: his different interviews as to where AI is going. 53 00:02:50,660 --> 00:02:54,920 Ejaaz: So he has some pretty sizable positions in some companies that a lot of people 54 00:02:54,920 --> 00:02:57,560 Ejaaz: may not have heard of purely because they're quite unsexy. 55 00:02:57,700 --> 00:03:02,560 Ejaaz: So he has an almost 9% position, so 10% position of the fund in this thing called 56 00:03:02,560 --> 00:03:09,600 Ejaaz: Astera Labs. Now, Astro Labs can be described as kind of like the connectivity layer between GPUs. 57 00:03:09,660 --> 00:03:14,680 Ejaaz: So if you imagine a data center, you have GPUs, they're kind of like the engine that can like pre-train 58 00:03:14,950 --> 00:03:18,450 Ejaaz: post-train your model and also like inference your model. But in order for these 59 00:03:18,450 --> 00:03:22,610 Ejaaz: GPUs to work, they need to traverse a bunch of different data. 60 00:03:22,770 --> 00:03:26,050 Ejaaz: They need to send data between themselves. They also need to access data from 61 00:03:26,050 --> 00:03:28,770 Ejaaz: all these memory chips that they store this data on. 62 00:03:29,030 --> 00:03:33,030 Ejaaz: Now, in order to access this, you need some kind of a plumbing system. 63 00:03:33,150 --> 00:03:36,010 Ejaaz: And I'm being very high level on this because I'm not going to pretend that 64 00:03:36,010 --> 00:03:37,830 Ejaaz: I know the intricacies of all of this. 65 00:03:38,190 --> 00:03:40,990 Ejaaz: Astero Labs is a company that essentially fixes that. 66 00:03:41,050 --> 00:03:45,430 Ejaaz: The problem it solves is as AI clusters scale to hundreds of thousands of chips, 67 00:03:45,830 --> 00:03:49,810 Ejaaz: The bottleneck stops becoming GPUs specifically, but it starts becoming that 68 00:03:49,810 --> 00:03:52,750 Ejaaz: transfer window and sending the right data at the right time, 69 00:03:52,930 --> 00:03:57,510 Ejaaz: accessing the right data at the right time is what the plumbing system that Astro Labs builds. 70 00:03:57,650 --> 00:04:01,590 Ejaaz: Now, I haven't heard of Astro Labs until we started researching for this episode, 71 00:04:01,590 --> 00:04:07,410 Ejaaz: but I remember another company being the exact same case called Cerebrus Labs, 72 00:04:07,530 --> 00:04:12,610 Ejaaz: which is what Gavin was talking about almost like, I think like six months ago, which is quite long, 73 00:04:12,730 --> 00:04:15,230 Ejaaz: given the relative scale or timeline of AI. 74 00:04:15,850 --> 00:04:18,470 Ejaaz: And then the next thing I know is that they IPO'd for like, what was it, 75 00:04:18,550 --> 00:04:21,950 Ejaaz: like $60 billion, and it's just been like, it's up 40% since the IPO. 76 00:04:22,190 --> 00:04:25,810 Ejaaz: So just kind of point towards these different trends. Asher Labs might be something 77 00:04:25,810 --> 00:04:27,150 Ejaaz: significant on that horizon. 78 00:04:27,590 --> 00:04:30,690 Josh: Yeah, Cerebrus is one of his earliest investments. I mean, he was in Cerebrus 79 00:04:30,690 --> 00:04:34,110 Josh: very early in the company's lifetime, which means he's bet on this thesis for years. 80 00:04:34,230 --> 00:04:36,750 Josh: There's also a few other ones that he's been betting on for a long time. 81 00:04:36,890 --> 00:04:40,130 Josh: I mean, NVIDIA being that flagship one, being involved in NVIDIA for 20 plus 82 00:04:40,130 --> 00:04:43,710 Josh: years is pretty incredible and still having conviction all the way through is impressive. 83 00:04:44,130 --> 00:04:48,210 Josh: Gavin was recently on two podcasts where I was listening to him speak about his NVIDIA position. 84 00:04:48,370 --> 00:04:50,850 Josh: And it's very clear that he believes that they're going to be able to maintain 85 00:04:50,850 --> 00:04:55,030 Josh: these profit margins and maintain the demand as well, which means he's putting 86 00:04:55,030 --> 00:04:59,350 Josh: NVIDIA on a clear path to getting close to $10 trillion in market cap. 87 00:04:59,450 --> 00:05:01,030 Josh: It's only halfway there right now. 88 00:05:01,150 --> 00:05:04,650 Josh: A few other noteworthy mentions are Micron, which we discussed on a previous 89 00:05:04,650 --> 00:05:07,890 Josh: episode, which I would highly recommend going to watch in terms of the AI investment 90 00:05:07,890 --> 00:05:09,770 Josh: stack and where all these companies lie. 91 00:05:10,230 --> 00:05:14,110 Josh: Micron is one of the largest memory makers and a crazy statistic about Micron. 92 00:05:14,330 --> 00:05:15,930 Josh: A year ago, it was sub $100 billion. 93 00:05:16,410 --> 00:05:21,090 Josh: And as of recording this, it just eclipsed a trillion dollars in market cap. 94 00:05:21,230 --> 00:05:26,810 Josh: It got a 10X in a single year. And it's a testament to how important that memory problem is. 95 00:05:27,210 --> 00:05:31,490 Josh: Now, perhaps some of the less noteworthy companies that are interesting, 96 00:05:31,590 --> 00:05:33,570 Josh: EGES, there's one I want to highlight for you in particular, 97 00:05:33,630 --> 00:05:36,430 Josh: because I feel like you're going to like this one, is Unity Software. 98 00:05:37,020 --> 00:05:41,640 Josh: And for those that don't know, Unity, I mean, I know this very well as a gamer, Unity is a game engine. 99 00:05:41,820 --> 00:05:45,360 Josh: There's a lot of popular games that are built using this 3D rendering software. 100 00:05:45,500 --> 00:05:49,360 Josh: So why would someone who's investing in AI be investing in Unity software, 101 00:05:49,480 --> 00:05:53,200 Josh: the thing that makes my video games? And the answer is the 3D game engine. 102 00:05:53,640 --> 00:05:58,760 Josh: Unity is a world model builder, and it has a really deep understanding of physics 103 00:05:58,760 --> 00:06:01,920 Josh: and the way the world works and the understanding of textures and lighting. 104 00:06:02,100 --> 00:06:06,420 Josh: And when these AI companies are trying to build AGI, they're trying to build humanoid robots. 105 00:06:06,680 --> 00:06:10,220 Josh: A big part of that is simulating virtual environments and virtual datasets that 106 00:06:10,220 --> 00:06:12,140 Josh: allow these robots to be trained in. 107 00:06:12,320 --> 00:06:15,320 Josh: And Unity just still happens to be one of the best ones. So I feel like that 108 00:06:15,320 --> 00:06:16,880 Josh: one I want to highlight specifically for you, EHS. 109 00:06:17,240 --> 00:06:22,000 Josh: As the world model Maxi, there is a clear path in which a gaming company that 110 00:06:22,000 --> 00:06:25,520 Josh: is known for a scanning engine becomes a pretty serious player in the world of AI. 111 00:06:25,800 --> 00:06:28,760 Ejaaz: Yeah, the whole thesis behind world models is pretty simple. 112 00:06:29,020 --> 00:06:34,480 Ejaaz: It's AI models or LLMs currently understand the world through text, through books. 113 00:06:34,620 --> 00:06:38,060 Ejaaz: It's kind of like a student sitting in a library, but it doesn't actually have 114 00:06:38,060 --> 00:06:39,500 Ejaaz: experience of the real world. 115 00:06:39,620 --> 00:06:43,020 Ejaaz: World models basically unlock that. It's like putting a game character into 116 00:06:43,020 --> 00:06:47,040 Ejaaz: a simulated environment and understanding the physical reality of how the world works. 117 00:06:47,120 --> 00:06:50,740 Ejaaz: If it works, if I drop this phone, if I kick a ball, like what happens? 118 00:06:50,820 --> 00:06:53,280 Ejaaz: What are the next consequential steps? What do you do? 119 00:06:53,720 --> 00:06:57,180 Ejaaz: World models effectively fix that. And there are very few players that have 120 00:06:57,180 --> 00:07:01,200 Ejaaz: like built this out at scale. I think currently the leader is probably Google with Genie 3. 121 00:07:01,400 --> 00:07:04,440 Ejaaz: They actually released a new model called Gemini Omni recently, 122 00:07:04,660 --> 00:07:08,200 Ejaaz: which kind of like does this at scale, but it's not quite like where it's meant 123 00:07:08,200 --> 00:07:10,300 Ejaaz: to be. It hasn't quite had its ChatGPT moment. 124 00:07:10,860 --> 00:07:14,660 Ejaaz: What I like about Gavin in particular is he kind of has this barbell. 125 00:07:14,840 --> 00:07:18,220 Ejaaz: I don't know if you noticed this, Josh, where he's kind of old school and he's like, 126 00:07:18,700 --> 00:07:21,240 Ejaaz: people are going to need GPUs. People are going to need memory. 127 00:07:21,560 --> 00:07:24,620 Ejaaz: I'm going to invest in the biggest players, Micron and NVIDIA. 128 00:07:24,860 --> 00:07:27,240 Ejaaz: But then he has this kind of forward-looking thing where he's like, 129 00:07:27,380 --> 00:07:30,000 Ejaaz: I think that's where the puck is going to go. 130 00:07:30,260 --> 00:07:35,020 Ejaaz: And so I think, let me invest in Cerebrus because I think inference is going to be super important. 131 00:07:35,260 --> 00:07:37,920 Ejaaz: And then let me invest in Unity because I think world models are going to be 132 00:07:37,920 --> 00:07:41,200 Ejaaz: the future of how we train robots and future LLMs. 133 00:07:41,340 --> 00:07:45,440 Ejaaz: So he has this kind of like barbell approach. Now, one thing that I see in his 134 00:07:45,440 --> 00:07:48,660 Ejaaz: portfolio over here as well is this, well, there's two companies, actually. 135 00:07:48,880 --> 00:07:52,800 Ejaaz: It's this company called Positron, which kind of creates inference chips. 136 00:07:52,920 --> 00:07:56,720 Ejaaz: Now, if that sounds similar, it's very similar to Cerebrus. That's exactly what they do. 137 00:07:56,920 --> 00:08:00,780 Ejaaz: And it's around this entire thesis, which Gavin has spoken about on his recent 138 00:08:00,780 --> 00:08:05,140 Ejaaz: interviews, which is the infrastructure stack, specifically the training stack 139 00:08:05,140 --> 00:08:10,560 Ejaaz: for AI models, is moving from pre-training to something more focused on post-training. 140 00:08:10,660 --> 00:08:14,080 Ejaaz: Now, if you've been involved in the AI sphere, generally, you kind of had an 141 00:08:14,080 --> 00:08:18,080 Ejaaz: idea that this shift was happening. But Gavin is all in on this thesis. 142 00:08:18,320 --> 00:08:24,080 Ejaaz: And if you have a model, it still needs to understand new information that comes in, new data. 143 00:08:24,200 --> 00:08:27,460 Ejaaz: It needs to update it. Just because you pre-train it on a specific data set 144 00:08:27,460 --> 00:08:29,360 Ejaaz: doesn't mean it's going to be a genius for the rest of its life. 145 00:08:29,460 --> 00:08:31,300 Ejaaz: It still needs to learn new information. 146 00:08:31,440 --> 00:08:34,840 Ejaaz: That happens in the post-training layer, and it requires a lot of inference. 147 00:08:35,320 --> 00:08:39,220 Ejaaz: Secondly, if you need the AI model to actually think about the problem more, 148 00:08:39,500 --> 00:08:41,960 Ejaaz: in the same way that we take in new information, we're like, 149 00:08:42,060 --> 00:08:45,980 Ejaaz: hmm, I wonder if this angle makes sense or if another thesis maybe applies. 150 00:08:46,140 --> 00:08:48,780 Ejaaz: That's known as reasoning. You need a lot of inference. Now, 151 00:08:48,880 --> 00:08:53,560 Ejaaz: the estimates are the cost or the revenue opportunity from inference alone is 152 00:08:53,560 --> 00:08:58,640 Ejaaz: worth around 5 to 10x more than the amount of compute that is being put into pre-training. 153 00:08:58,820 --> 00:09:02,460 Ejaaz: So, AI labs and chip makers are suddenly making this big shift. 154 00:09:02,460 --> 00:09:07,440 Ejaaz: Like you've seen NVIDIA create a bunch of different GPUs that are aligned to 155 00:09:07,440 --> 00:09:09,840 Ejaaz: inference to allow agentic kind of exposure. 156 00:09:10,020 --> 00:09:13,320 Ejaaz: And so we're seeing Gavin express this through his different investments on inference alone. 157 00:09:13,580 --> 00:09:17,280 Ejaaz: And the final point that I'll make is Gavin made a really cool point, 158 00:09:17,400 --> 00:09:21,040 Ejaaz: and we're talking about this before we started recording, Josh, on China specifically. 159 00:09:21,420 --> 00:09:24,620 Ejaaz: It's been very much like China versus the US when it comes to the AI race. 160 00:09:25,000 --> 00:09:29,300 Ejaaz: China has a very unique kind of setup where they have an abundance of energy 161 00:09:29,300 --> 00:09:32,980 Ejaaz: and the ability to scale chip manufacturers. That's something that the U.S. 162 00:09:33,120 --> 00:09:38,060 Ejaaz: Currently struggles with. That's why they outsource a ton of stuff to the likes of TSMC on Taiwan. 163 00:09:38,300 --> 00:09:42,260 Ejaaz: And what he basically explains is that China has a unique opportunity to create 164 00:09:42,260 --> 00:09:46,860 Ejaaz: infrastructure or chips specifically that are going to look very different to what the U.S. 165 00:09:46,920 --> 00:09:49,400 Ejaaz: Is creating because they're focused so much on inference. 166 00:09:49,600 --> 00:09:53,400 Ejaaz: So Gavin, you could say, is leading the charge in the U.S. through his investments 167 00:09:53,400 --> 00:09:56,260 Ejaaz: on building or taking a bet on the U.S. 168 00:09:56,400 --> 00:10:00,100 Ejaaz: Infrastructure setup for inference. And I think that could be a huge opportunity in the future. 169 00:10:00,100 --> 00:10:03,320 Josh: And it's worth noting that this bet also isn't only for the upside. 170 00:10:03,500 --> 00:10:07,580 Josh: There is a large put position here in an ETF named QQQ. 171 00:10:07,740 --> 00:10:11,820 Josh: Now, for those who are not familiar, it covers the NASDAQ 100. It's a basket of stocks. 172 00:10:11,960 --> 00:10:17,580 Josh: It's the second most traded ETF in the United States. And it's been performing incredibly. 173 00:10:17,900 --> 00:10:23,900 Josh: In 2023, it was up 55%. 2024, 25%. 2025, 20%. And so far in 2026, 174 00:10:24,400 --> 00:10:29,900 Josh: already up 17%. So QQQ, as an index fund, has been doing incredibly well. 175 00:10:30,020 --> 00:10:33,180 Josh: It's easy. It's a basket of the top 100 stocks. Gavin is saying, 176 00:10:33,820 --> 00:10:37,440 Josh: I'm shorting against that. I think that's going to go down. I think you're wrong. 177 00:10:37,600 --> 00:10:42,180 Josh: And what this tells me is that he believes strongly in the AI play in the sense 178 00:10:42,180 --> 00:10:45,260 Josh: that he's going to invest in the key makers who are solving these bottlenecks. 179 00:10:45,420 --> 00:10:49,420 Josh: But as a market-wide general sentiment, he doesn't appear to be very bullish. 180 00:10:49,520 --> 00:10:54,440 Josh: And this is a hedge against that downside protection, where if the market does 181 00:10:54,440 --> 00:10:58,620 Josh: start to fall apart in ways that are less favorable, even though AI still wins, 182 00:10:58,780 --> 00:11:00,220 Josh: he has that hedge with QQQ. 183 00:11:01,270 --> 00:11:05,050 Josh: We can kind of break this down into a few bottlenecks that he believes are going 184 00:11:05,050 --> 00:11:07,950 Josh: to be most important when it comes to investing. 185 00:11:08,170 --> 00:11:12,470 Josh: These key things that he is looking at in terms of what the world of AI is going 186 00:11:12,470 --> 00:11:15,050 Josh: to need as we progress forward. What are the actual constraints? 187 00:11:15,330 --> 00:11:17,970 Josh: What do they look like? And then how do you invest? How do you convert these 188 00:11:17,970 --> 00:11:21,010 Josh: into actual dollars that you could put into companies to earn you money? 189 00:11:21,150 --> 00:11:22,130 Josh: And there's four of them. 190 00:11:22,450 --> 00:11:25,590 Josh: The first one is verticalized small language models. 191 00:11:25,710 --> 00:11:30,150 Ejaaz: If you think about LLMs in general, like the chatbots that you talk to, 192 00:11:30,150 --> 00:11:33,370 Ejaaz: such as Claude and ChatGPT, they're generalized LLM. 193 00:11:33,510 --> 00:11:37,170 Ejaaz: So they have a wide understanding of the world in context, and they'll be able 194 00:11:37,170 --> 00:11:38,610 Ejaaz: to answer like specific questions. 195 00:11:38,810 --> 00:11:43,750 Ejaaz: But it's another thing, training a model around a specific vertical or specific 196 00:11:43,750 --> 00:11:46,770 Ejaaz: problem that you're trying to solve. Where do these specific problems exist? 197 00:11:46,970 --> 00:11:50,770 Ejaaz: Well, in enterprises that are deep on solving a particular problem or companies 198 00:11:50,770 --> 00:11:54,110 Ejaaz: that have made or formed a niche in their particular subsector. 199 00:11:54,550 --> 00:11:58,270 Ejaaz: Now, verticalized small language models basically address exactly this. 200 00:11:58,370 --> 00:12:01,910 Ejaaz: They're frontier models, but highly optimized towards running efficiently on 201 00:12:01,910 --> 00:12:04,130 Ejaaz: specific enterprise data or locally on device. 202 00:12:04,290 --> 00:12:07,570 Ejaaz: Now, we have spoken about on device or locally run models before, 203 00:12:07,970 --> 00:12:11,850 Ejaaz: purely for the case that there's a bunch of data in your phone or the devices 204 00:12:11,850 --> 00:12:16,210 Ejaaz: that you use that are highly personable to you, but you don't necessarily want to give up. 205 00:12:16,330 --> 00:12:19,650 Ejaaz: And companies don't necessarily have access to that. For example, 206 00:12:20,110 --> 00:12:21,950 Ejaaz: medical records, financial details. 207 00:12:22,210 --> 00:12:26,650 Ejaaz: I saw OpenAI release a financial AI agent that can get access to your bank account, 208 00:12:26,650 --> 00:12:30,730 Ejaaz: But it can't actually act on it because there's a lot of personal identifiable 209 00:12:30,730 --> 00:12:34,310 Ejaaz: information that you don't want to share, such as your social security number, banking details, etc. 210 00:12:34,790 --> 00:12:38,770 Ejaaz: Now, locally run models or these SLMs can solve that kind of a problem. 211 00:12:38,950 --> 00:12:42,250 Ejaaz: And Gavin is making a huge bet that these are going to become huge in the future. 212 00:12:42,730 --> 00:12:46,850 Ejaaz: One company that I've noted that Gavin is hugely bullish on is Apple. 213 00:12:46,970 --> 00:12:50,150 Ejaaz: Although he doesn't express an investment interest, he knows or thinks that 214 00:12:50,150 --> 00:12:53,390 Ejaaz: Apple is going to be the device maker, one of the major device makers, 215 00:12:54,040 --> 00:12:57,380 Ejaaz: Allows for these locally run models to run on their devices. 216 00:12:57,700 --> 00:13:01,420 Ejaaz: Now, in a world where that is the future, you can start thinking of a world 217 00:13:01,420 --> 00:13:05,700 Ejaaz: where maybe Claw doesn't need to be the model that you need to interface with every day. 218 00:13:05,820 --> 00:13:08,740 Ejaaz: Maybe you need a personalized AI agent trained on your own data, 219 00:13:08,760 --> 00:13:11,560 Ejaaz: and that's what these SLMs end up eventually becoming. 220 00:13:11,660 --> 00:13:15,340 Ejaaz: Now, that's the generalized version of it where you can run it on your own phone, 221 00:13:15,540 --> 00:13:19,400 Ejaaz: but a bunch of enterprises will run these highly optimized and specialized models 222 00:13:19,400 --> 00:13:23,360 Ejaaz: to train on their proprietary data, which ends up helping them sell a product 223 00:13:23,360 --> 00:13:25,060 Ejaaz: or market a product much better. 224 00:13:25,700 --> 00:13:29,440 Josh: Oh man, Apple's in such a good position there. So good. I can't wait for WWDC. It's coming. 225 00:13:29,740 --> 00:13:32,520 Josh: We are just a few weeks away from Apple's developer conference where they're 226 00:13:32,520 --> 00:13:36,600 Josh: going to unveil all of this new AI software that's coming and what that looks 227 00:13:36,600 --> 00:13:37,480 Josh: like integrating the hardware. 228 00:13:37,620 --> 00:13:40,080 Josh: That's going to be huge. We will be covering that. I'm so excited to talk about 229 00:13:40,080 --> 00:13:44,100 Josh: that. In terms of the next pillar of this, it is sovereign infrastructure. 230 00:13:44,100 --> 00:13:48,940 Josh: We always talk about this, that the speed of bits is so much faster than the speed of atoms. 231 00:13:49,120 --> 00:13:53,800 Josh: I mean, When you think about AI infrastructure, the quality of models has gone purely exponential. 232 00:13:54,100 --> 00:13:57,200 Josh: The amount of intelligence we could generate per watt, the amount of intelligence 233 00:13:57,200 --> 00:13:59,480 Josh: per token, is up into the right only. 234 00:13:59,780 --> 00:14:03,400 Josh: What isn't up into the right at nearly the same rate is the speed of physical 235 00:14:03,400 --> 00:14:05,640 Josh: deployment, because that itself is the mode. 236 00:14:05,760 --> 00:14:09,200 Josh: It's very difficult to take hardware that is incredibly complicated. 237 00:14:09,340 --> 00:14:12,720 Josh: We're talking about transistors that are down to the atomic level of precision 238 00:14:12,720 --> 00:14:16,880 Josh: and deploying them at scale in a world in which our infrastructure is already 239 00:14:16,880 --> 00:14:20,380 Josh: suffering. I mean, with the acceleration of electric cars, the grids have already 240 00:14:20,380 --> 00:14:23,700 Josh: been feeling a little bit more restrained. They're kind of at max capacity. 241 00:14:23,960 --> 00:14:26,220 Josh: Now they have the energy problem. Now they have the chip problem. 242 00:14:26,680 --> 00:14:30,540 Josh: Gavin is very strongly betting on the fact that infrastructure is hard. 243 00:14:30,740 --> 00:14:34,220 Josh: It's going to take many, many days to months to years to do. 244 00:14:34,340 --> 00:14:36,760 Josh: And he's betting on the people that can compress that into weeks. 245 00:14:36,900 --> 00:14:39,600 Josh: So the speed of the physical deployment itself is the moat. 246 00:14:39,800 --> 00:14:43,740 Josh: He's kind of narrowing in on who the companies are that are able to deploy as 247 00:14:43,740 --> 00:14:46,680 Josh: fast as possible. When I think about this, my first thought is SpaceX. 248 00:14:46,680 --> 00:14:49,760 Josh: And how quickly they've been able to build Colossus and then rent that out to 249 00:14:49,760 --> 00:14:52,440 Josh: Anthropic and I'm sure companies in the future. 250 00:14:52,660 --> 00:14:56,680 Josh: But that infrastructure pillar is one of the key ones that he's looking at. 251 00:14:56,720 --> 00:14:57,820 Josh: I think everyone, we looked at. 252 00:14:58,410 --> 00:15:01,470 Josh: Leopold's portfolio as well, that was a core component of that. 253 00:15:01,590 --> 00:15:04,130 Josh: It's just, it's really hard to build things. And whoever can build things, 254 00:15:04,530 --> 00:15:05,650 Josh: they can sell it for a lot of money. 255 00:15:05,770 --> 00:15:09,750 Josh: SpaceX, their largest line item now in terms of revenue is the data center that 256 00:15:09,750 --> 00:15:11,390 Josh: they're renting out. It has nothing to do with rocket ships. 257 00:15:11,450 --> 00:15:13,610 Josh: And I think that's a testament to how important this pillar is. 258 00:15:13,930 --> 00:15:16,810 Ejaaz: So it's speed that he cares about and he thinks that it's important, 259 00:15:16,930 --> 00:15:18,990 Ejaaz: but it's also like the cost, right? 260 00:15:19,110 --> 00:15:24,530 Ejaaz: He keeps referencing this metric, which is performance per watt. 261 00:15:24,530 --> 00:15:25,430 Josh: Perf per watt. 262 00:15:25,610 --> 00:15:29,750 Ejaaz: Yeah, perf per watt. And what he's talking about here is companies are increasingly, 263 00:15:30,030 --> 00:15:34,770 Ejaaz: companies being AI labs, are increasingly caring about how many tokens can you 264 00:15:34,770 --> 00:15:36,770 Ejaaz: generate per watt, right? 265 00:15:36,910 --> 00:15:40,610 Ejaaz: Because if you think about spending billions and trillions of dollars, which is what... 266 00:15:41,110 --> 00:15:46,730 Ejaaz: Five companies are currently spending this year alone on GPUs and compute to 267 00:15:46,730 --> 00:15:49,630 Ejaaz: kind of like power these things or electricity to power these things, 268 00:15:49,910 --> 00:15:53,470 Ejaaz: You want to get a lot of bang for buck, especially if you're scaling to the 269 00:15:53,470 --> 00:15:55,710 Ejaaz: size that most of these hyperscalers are doing. 270 00:15:55,910 --> 00:16:01,970 Ejaaz: So if you think about it, if I prompt Claude and if I prompt ChatGPT and Claude 271 00:16:01,970 --> 00:16:06,910 Ejaaz: gives me an answer that cost me two cents and ChatGPT gives me an answer that cost me $1, 272 00:16:06,990 --> 00:16:11,110 Ejaaz: I'm probably going to end up using Claude, even if it's like hypothetically 273 00:16:11,110 --> 00:16:16,950 Ejaaz: like say 95% of the intelligence hypothetically that ChatGPT has. 274 00:16:17,090 --> 00:16:18,090 Ejaaz: Because the point is you can prompt 275 00:16:18,090 --> 00:16:20,990 Ejaaz: it even more and eventually get to the answer for a lot less of cost. 276 00:16:21,130 --> 00:16:24,650 Ejaaz: So cost becomes a huge thing. This week alone, Microsoft and Uber announced 277 00:16:24,650 --> 00:16:28,790 Ejaaz: that they are effectively reducing their exposure to clawed code because their 278 00:16:28,790 --> 00:16:32,010 Ejaaz: annual budgets kind of like got sequestered in about four months. 279 00:16:32,110 --> 00:16:37,730 Ejaaz: So the point is like the cost of getting access to this intelligence matters a lot. 280 00:16:37,730 --> 00:16:41,110 Ejaaz: And you see this across Gavin's investment portfolio with Cerebras, 281 00:16:41,250 --> 00:16:44,790 Ejaaz: with Alexa, Positron, and with Astro Labs, which basically, like, 282 00:16:44,850 --> 00:16:49,970 Ejaaz: what I've noticed is he identifies these really niche infrastructure bottlenecks, 283 00:16:50,010 --> 00:16:51,590 Ejaaz: and he basically makes one simple bet. 284 00:16:51,670 --> 00:16:56,150 Ejaaz: He's like, yeah, if this company solves that, the performance per watt gets 285 00:16:56,150 --> 00:16:59,670 Ejaaz: to this specific level, which means that AI Labs are probably going to buy more 286 00:16:59,670 --> 00:17:02,070 Ejaaz: of these GPUs or more of these companies or more of these things, 287 00:17:02,130 --> 00:17:06,730 Ejaaz: and you end up with a bottleneck being resolved by one of these different companies that he's betting on. 288 00:17:06,730 --> 00:17:10,530 Ejaaz: So his thesis actually is quite simple, although it's quite complex, 289 00:17:10,670 --> 00:17:14,010 Ejaaz: which is I'm just going to focus on the AI bottlenecks at the infrastructure level. 290 00:17:14,130 --> 00:17:17,490 Ejaaz: And if I can identify a company that can increase performance for what by this 291 00:17:17,490 --> 00:17:21,570 Ejaaz: amount, that can make tokens this amount cheaper, then my bet is those companies 292 00:17:21,570 --> 00:17:25,250 Ejaaz: are going to be valuable and will either IPO or get acquired for a large sum of money. 293 00:17:25,690 --> 00:17:28,670 Josh: And the thinkers to know for the section, for those looking to copy trade, 294 00:17:28,830 --> 00:17:33,270 Josh: Gavin, Estera Labs, we have Cerebrus, we have Sci-5, and we have Positron. 295 00:17:33,490 --> 00:17:36,250 Josh: Those are the four companies that are really critical in this sector. 296 00:17:36,370 --> 00:17:39,930 Josh: Now, the fourth and final is a combination of energy and space. 297 00:17:40,170 --> 00:17:42,530 Josh: Because, I mean, like we talked about in the previous point, 298 00:17:42,850 --> 00:17:45,250 Josh: the terrestrial grid very much limits energy. 299 00:17:45,410 --> 00:17:50,490 Josh: And it's very difficult to earn new energy. I think a statistic like 40% of 300 00:17:50,490 --> 00:17:54,470 Josh: new data centers have very strong petitions against them, or lobbying against 301 00:17:54,470 --> 00:17:57,090 Josh: them, people protesting they do not want these data centers. 302 00:17:57,730 --> 00:18:02,130 Josh: There's a lot of resistance to them. And the way that we solve this is twofold. 303 00:18:02,350 --> 00:18:06,030 Josh: One is creating energy out of the box. It's kind of portable energy you could 304 00:18:06,030 --> 00:18:07,850 Josh: bring to these data centers, power them with a smaller phone. 305 00:18:08,450 --> 00:18:11,310 Josh: That's companies like Bloom Energy, who Leopold is very bullish on. 306 00:18:11,490 --> 00:18:14,690 Josh: But then there's also the orbital compute part of this, which is what Gavin 307 00:18:14,690 --> 00:18:16,510 Josh: has really shifted his focus to. 308 00:18:16,650 --> 00:18:19,910 Josh: Now, the first largest company in this sector, of course, is SpaceX. 309 00:18:20,270 --> 00:18:23,290 Josh: They're the only one who is capable of being the highway to space, 310 00:18:23,290 --> 00:18:28,230 Josh: to delivering actual payload to orbit, to getting maths and data centers in 311 00:18:28,230 --> 00:18:32,210 Josh: low-end orbit that can generate enough intelligence and power that can funnel this. 312 00:18:32,370 --> 00:18:34,890 Josh: I think the space stack is a little bit bigger than just SpaceX. 313 00:18:35,230 --> 00:18:40,090 Josh: I was surprised to see there wasn't more allocation of space stocks in his portfolio, 314 00:18:40,270 --> 00:18:43,030 Josh: given the fact that he believes this is such a huge industry. 315 00:18:43,350 --> 00:18:47,810 Josh: Perhaps the reality is that it's just too early and that SpaceX is the linchpin 316 00:18:47,810 --> 00:18:52,090 Josh: to unlocking this industry and just kind of closely evaluating Starship V3 launches. 317 00:18:52,090 --> 00:18:54,590 Josh: We had a Starship launch last week. It performed very well. 318 00:18:55,050 --> 00:18:59,650 Josh: Without Starship functioning, we don't get energy in space. We don't get racks 319 00:18:59,650 --> 00:19:02,550 Josh: to orbit. It is required because the amount of payload is so large. 320 00:19:02,670 --> 00:19:06,130 Josh: So I'm sure SpaceX is the one to watch. There's a lot of other second order 321 00:19:06,130 --> 00:19:07,730 Josh: companies that can be impacted by this. 322 00:19:08,540 --> 00:19:12,200 Josh: I think we want to get to the end of this with the question that a lot of people 323 00:19:12,200 --> 00:19:15,840 Josh: are going to be asking, which is why is this not just the dot-com bubble again? 324 00:19:16,300 --> 00:19:19,960 Josh: And Gavin was asked this question many times. He has some pretty strong answers 325 00:19:19,960 --> 00:19:24,620 Josh: and I kind of believe him. It seemed his case that he made was pretty convincing. 326 00:19:25,160 --> 00:19:29,120 Ejaaz: Okay, so the way I think of his case is as follows. 327 00:19:29,840 --> 00:19:34,360 Ejaaz: In the dot-com bubble in the 2000s, it was debt-fueled. 328 00:19:34,480 --> 00:19:40,000 Ejaaz: You had people borrowing a hell of a lot of money for unproven theses and products 329 00:19:40,000 --> 00:19:42,160 Ejaaz: that people didn't actually care about or use. 330 00:19:42,460 --> 00:19:46,500 Ejaaz: Now, if you compare this to the current AI super cycle, as he describes it, 331 00:19:47,040 --> 00:19:50,820 Ejaaz: just from OpenAI and Anthropic alone, they're on track to reach $200 billion 332 00:19:50,820 --> 00:19:54,120 Ejaaz: of ARR this year, just two companies. 333 00:19:54,480 --> 00:19:58,560 Ejaaz: And this isn't made up money. This is money that they've signed through contracts 334 00:19:58,560 --> 00:20:02,900 Ejaaz: that are already prepaid in large part, I think 40 to 60% from a bunch of enterprise 335 00:20:02,900 --> 00:20:06,980 Ejaaz: and retail customers that are funding this. So this is real money exchanging hands. 336 00:20:07,180 --> 00:20:10,960 Ejaaz: Now, if you look at the GPU computer, so let's not look at the model labs, 337 00:20:11,060 --> 00:20:14,380 Ejaaz: let's look at the infrastructure, the people who are buying the goods from NVIDIA. 338 00:20:15,160 --> 00:20:21,000 Ejaaz: Google, Microsoft, Amazon, and Meta are all paying from their own cash reserves. 339 00:20:21,020 --> 00:20:23,960 Ejaaz: So they also haven't borrowed any money at all. 340 00:20:24,100 --> 00:20:26,160 Ejaaz: They're just spending free cash flow on this. 341 00:20:26,440 --> 00:20:28,820 Ejaaz: Amazon just came to the end of their free cash flow. Now, if they start borrowing 342 00:20:28,820 --> 00:20:32,540 Ejaaz: money, then we can start to get worried. But the point is, they are not levered up either. 343 00:20:32,760 --> 00:20:39,000 Ejaaz: Also, this is like five of the top companies in the entire world who are arguably 344 00:20:39,000 --> 00:20:41,540 Ejaaz: like some of the smartest companies because they are where they are in terms 345 00:20:41,540 --> 00:20:43,180 Ejaaz: of like market cap, value and size. 346 00:20:43,380 --> 00:20:48,080 Ejaaz: So the fact is like back to the dotcom bubble, you had a bunch of like no-name 347 00:20:48,080 --> 00:20:50,980 Ejaaz: companies that had raised a ton of money that was spending money in ways that 348 00:20:50,980 --> 00:20:52,020 Ejaaz: didn't really make sense. 349 00:20:52,520 --> 00:20:56,180 Ejaaz: In this cycle, you have like some of the smartest companies in the world spending 350 00:20:56,180 --> 00:20:59,940 Ejaaz: money that isn't levered at all, that isn't kind of like compressed into like 351 00:20:59,940 --> 00:21:01,060 Ejaaz: various different margins. 352 00:21:01,060 --> 00:21:04,080 Ejaaz: And all the quarterly reports that we've spoken about on previous episodes very 353 00:21:04,080 --> 00:21:07,540 Ejaaz: recently over the last couple of weeks have proven that profit is optimizing 354 00:21:07,540 --> 00:21:10,200 Ejaaz: through a bunch of these different movements that models are progressing, 355 00:21:10,340 --> 00:21:11,240 Ejaaz: they're getting more intelligent. 356 00:21:11,440 --> 00:21:15,540 Ejaaz: So the single argument that Gavin has is, this isn't a dot com bubble, 357 00:21:15,680 --> 00:21:19,120 Ejaaz: because we're not levered on money, but also because the bottlenecks themselves 358 00:21:19,120 --> 00:21:22,100 Ejaaz: that we're speaking about is constrained by physical atoms. 359 00:21:22,100 --> 00:21:25,080 Ejaaz: It's one thing being like, OK, I'm going to buy a bunch of memory chips and 360 00:21:25,080 --> 00:21:27,380 Ejaaz: I'm going to buy a bunch of GPUs. 361 00:21:27,460 --> 00:21:31,880 Ejaaz: But it's not like NVIDIA can oversell GPUs. It's not like Micron can oversell 362 00:21:31,880 --> 00:21:36,260 Ejaaz: AI memory chips because they just don't have the chip production facilities to be able to do that. 363 00:21:36,520 --> 00:21:40,200 Ejaaz: And so his simple argument is it's not a bubble if you can't oversupply the entire market. 364 00:21:40,320 --> 00:21:43,560 Ejaaz: We are constrained by the fact that we don't have enough picks and shovels to 365 00:21:43,560 --> 00:21:45,000 Ejaaz: do the thing. And that's what he's investing in. 366 00:21:45,390 --> 00:21:48,150 Josh: And there's one great point. If you scroll down just a hair on this artifact, 367 00:21:48,350 --> 00:21:51,070 Josh: you can see the $2 to $3 trillion number. 368 00:21:51,330 --> 00:21:54,970 Josh: This is something that Gavin believes is a reality, that NVIDIA could sell $2 369 00:21:54,970 --> 00:22:01,070 Josh: to $3 trillion of GPUs this year and next year if only TSMC could supply them. 370 00:22:01,330 --> 00:22:07,390 Josh: So he's saying that TSMC is actually one of the major linchpins of bubble territory. 371 00:22:07,630 --> 00:22:09,950 Josh: And I'll explain why for a second, because I found this really interesting. 372 00:22:10,230 --> 00:22:15,210 Josh: If TSMC were able to supply the amount of demand that is required from these 373 00:22:15,210 --> 00:22:19,290 Josh: companies to actually provide them with that many chips, it will cost them a 374 00:22:19,290 --> 00:22:20,790 Josh: tremendous amount of money. 375 00:22:20,910 --> 00:22:23,610 Josh: And in fact, EJS, if you scroll up just a little bit more, you could see the 376 00:22:23,610 --> 00:22:25,870 Josh: earned income to the CapEx. 377 00:22:26,050 --> 00:22:27,910 Josh: There's like this kind of bar chart there. Yeah, right here, 378 00:22:28,030 --> 00:22:32,290 Josh: where you could see AI CapEx is not much different than the operating cash. 379 00:22:32,450 --> 00:22:36,010 Josh: And so far, companies are generating enough cash to fund the build out of these things. 380 00:22:36,130 --> 00:22:39,770 Josh: In the case that TSMC came to NVIDIA tomorrow and said, actually, 381 00:22:39,790 --> 00:22:41,490 Josh: we could triple our capacity overnight. 382 00:22:41,890 --> 00:22:45,210 Josh: NVIDIA wouldn't say no. And they would start spending an unbelievable amount 383 00:22:45,210 --> 00:22:46,510 Josh: of money to buy those chips. 384 00:22:46,790 --> 00:22:51,270 Josh: Other companies would then have to borrow money to fund the purchase of all these chips. 385 00:22:51,550 --> 00:22:55,530 Josh: And therefore, we would start to see that CapEx bubble really start to grow 386 00:22:55,530 --> 00:22:57,710 Josh: and separate from the operating cash of these companies. 387 00:22:57,870 --> 00:23:00,170 Josh: But because there are these supply constraints across the board, 388 00:23:00,230 --> 00:23:02,770 Josh: we have them in memory, we have them in chip making, we have them in energy, 389 00:23:03,150 --> 00:23:09,270 Josh: particularly as it relates to TSMC with the chips, we're not actually able to build out that fast. 390 00:23:09,510 --> 00:23:14,610 Josh: And therefore, TSMC is creating a blocker on the rate of acceleration of this bubble. 391 00:23:14,750 --> 00:23:18,130 Josh: And so long as TSMC stays limited and constrained to the amount of chips that 392 00:23:18,130 --> 00:23:22,650 Josh: they could produce, and so long as companies like Samsung and other chip makers don't. 393 00:23:23,240 --> 00:23:26,560 Josh: Actually overtake that market share, which seems improbable because it's going 394 00:23:26,560 --> 00:23:28,780 Josh: to take an incredibly long time and it's very difficult to do, 395 00:23:29,160 --> 00:23:32,600 Josh: then we're at a pretty sustainable rate of growth where it feels fast, 396 00:23:32,880 --> 00:23:36,100 Josh: but there's still this overwhelming amount of demand that can't be satiated 397 00:23:36,100 --> 00:23:37,780 Josh: because we just can't build it out fast enough. 398 00:23:38,040 --> 00:23:41,980 Josh: And so long as that dynamic stays intact, I think we're probably good for now. 399 00:23:42,120 --> 00:23:45,420 Ejaaz: Well, there's also this other thing, right? Because like you could assume that 400 00:23:45,420 --> 00:23:48,460 Ejaaz: like demand stays static, but that also doesn't happen. 401 00:23:48,600 --> 00:23:52,880 Ejaaz: You have an exponentially increasing demand side for all this AI stuff, 402 00:23:52,980 --> 00:23:57,820 Ejaaz: which is outpacing the production supply that we have for all of these different chips. 403 00:23:58,040 --> 00:24:04,220 Ejaaz: So the only way that I see this thesis kind of being unproven is if somehow 404 00:24:04,220 --> 00:24:11,000 Ejaaz: someone miraculously recreates ASML and we suddenly have ASML competitors all around. 405 00:24:11,140 --> 00:24:15,420 Ejaaz: By the way, those of you who don't know, ASML produces these $400 million machines, 406 00:24:15,620 --> 00:24:18,740 Ejaaz: which basically tsmc and every major chip fab 407 00:24:18,740 --> 00:24:21,440 Ejaaz: manufacturer needs and asml is only 408 00:24:21,440 --> 00:24:24,560 Ejaaz: one team in i believe norway that creates 409 00:24:24,560 --> 00:24:27,620 Ejaaz: these things and it takes ages and they're backlogged for like five 410 00:24:27,620 --> 00:24:33,020 Ejaaz: years right now or if we recreate a different type of llm that doesn't require 411 00:24:33,020 --> 00:24:36,880 Ejaaz: as many gpus or doesn't require as much memory but we're just simply not seeing 412 00:24:36,880 --> 00:24:42,180 Ejaaz: that i saw a story break today about sk hynix which is the number one memory 413 00:24:42,180 --> 00:24:46,820 Ejaaz: manufacturer and supplier for nvidia gpus they are basically the top dog when it comes to AI memory. 414 00:24:47,380 --> 00:24:50,760 Ejaaz: And they're currently courting, I think it's like $50 billion, 415 00:24:51,520 --> 00:24:56,720 Ejaaz: $50 to $100 billion worth of offers from the likes of Google and Microsoft. 416 00:24:56,920 --> 00:25:01,660 Ejaaz: So two companies alone, just to pay for the equipment that they need to scale 417 00:25:01,660 --> 00:25:06,580 Ejaaz: up future supply over the next three years to lock in supply that they're going 418 00:25:06,580 --> 00:25:07,820 Ejaaz: to create three years from now. 419 00:25:07,960 --> 00:25:11,000 Ejaaz: This is how desperate some of these big companies are for memory. 420 00:25:11,220 --> 00:25:14,020 Ejaaz: This is one subsector of the AI component. 421 00:25:14,480 --> 00:25:19,580 Ejaaz: And SK Hanox is instead just saying, no, I don't want to give you guarantees of our supply. 422 00:25:19,980 --> 00:25:23,560 Ejaaz: Instead, we'll just hike up the price. They have like 70% operating margin, 423 00:25:23,680 --> 00:25:26,860 Ejaaz: by the way, which is just unheard of in semiconductor companies at all. 424 00:25:26,960 --> 00:25:31,160 Ejaaz: So it makes complete sense why Gavin is just going all in, because it doesn't look like a bubble. 425 00:25:31,260 --> 00:25:34,480 Ejaaz: It might seem like a bubble. The market might react to it. We opened our stocks 426 00:25:34,480 --> 00:25:38,140 Ejaaz: portfolio today before we started recording and everything was like down. 427 00:25:38,260 --> 00:25:41,760 Ejaaz: But it's just reactionary because the directional goal of all of this is that 428 00:25:41,760 --> 00:25:45,460 Ejaaz: we're just going to need more GPUs. going to need more semiconductor chips and 429 00:25:45,460 --> 00:25:48,480 Ejaaz: we don't have enough supply we don't have enough manufacturers for this 430 00:25:49,190 --> 00:25:53,730 Josh: So in conclusion, watts and wafers, that's it. Those are the two brick walls. 431 00:25:53,870 --> 00:25:57,070 Josh: Those are the two limiting factors, the limiting constraints that are going 432 00:25:57,070 --> 00:25:59,830 Josh: to prevent us from accelerating too fast. 433 00:26:00,050 --> 00:26:03,710 Josh: And so long as those watts and wafers stay valuable, stay in high demand and 434 00:26:03,710 --> 00:26:07,030 Josh: stay limited in their supply, there will be good times ahead. 435 00:26:07,270 --> 00:26:11,490 Josh: Now, if you are looking for a TLDR on Gavin's portfolio, I will read out some 436 00:26:11,490 --> 00:26:12,550 Josh: of the largest holdings for you. 437 00:26:12,770 --> 00:26:15,330 Josh: That way you can just, again, not financial advice. This is what Gavin's in. 438 00:26:15,430 --> 00:26:16,290 Josh: This is not what we're in. 439 00:26:16,610 --> 00:26:19,070 Josh: I have no idea if any of these are gonna go up, down or in circles. 440 00:26:19,190 --> 00:26:22,590 Josh: But his largest position, ironically, is that QQQ put position. 441 00:26:22,830 --> 00:26:26,490 Josh: He's generally speaking bearish on the market, which I think is very noteworthy to talk about. 442 00:26:26,950 --> 00:26:33,210 Josh: Second to that is Astera Labs, 7.4%. Alab is the ticker. Third is Unity, the 3D software. 443 00:26:33,690 --> 00:26:36,890 Josh: Then there's a whole bunch of others. Sienna, Micron, NVIDIA, 444 00:26:37,030 --> 00:26:41,830 Josh: Amazon, Lumentum, Alphabet, Coherent, Roblox, EchoStar, Twilio, Wayfair. 445 00:26:42,110 --> 00:26:46,530 Josh: Wayfair is a furniture company. This guy's in everything. There's a lot of investments. 446 00:26:46,530 --> 00:26:49,690 Josh: I think if you're interested, you can take a look at a 13F online. 447 00:26:50,050 --> 00:26:52,790 Josh: We can link to it in the description. But that is the Gavin thesis, 448 00:26:52,970 --> 00:26:54,730 Josh: that the bottleneck is Watson wafers. 449 00:26:54,990 --> 00:26:59,010 Josh: So long as these things stay intact, we are up only. Ejaz, how are you ingesting 450 00:26:59,010 --> 00:27:01,050 Josh: this information and what are you doing with it? 451 00:27:01,370 --> 00:27:05,630 Ejaaz: So the market's been pretty rocky since the Leopold 13F. 452 00:27:05,750 --> 00:27:09,490 Ejaaz: And I'm starting to realize as we're recording this episode that Gavin's kind 453 00:27:09,490 --> 00:27:11,490 Ejaaz: of like an older, wiser Leopold. 454 00:27:11,590 --> 00:27:16,150 Ejaaz: He's been around for a while. He may not have $13 billion in AUM, 455 00:27:16,290 --> 00:27:18,790 Ejaaz: but I have a feeling he's going to be around a decade from now. 456 00:27:18,910 --> 00:27:21,670 Ejaaz: And so if you're listening to this and you're like, listen, I don't want to 457 00:27:21,670 --> 00:27:27,270 Ejaaz: keep up to date with AI every single minute of every single hour, every single day. 458 00:27:27,430 --> 00:27:31,170 Ejaaz: And I want to just kind of park my money and kind of like see how it grows over 459 00:27:31,170 --> 00:27:33,430 Ejaaz: the next couple of months or years. 460 00:27:33,830 --> 00:27:36,050 Ejaaz: Gavin's portfolio probably kind of makes a lot of sense. Again, 461 00:27:36,110 --> 00:27:38,630 Ejaaz: this is not investment advice, but he takes a more cautious, 462 00:27:39,030 --> 00:27:42,910 Ejaaz: long-term, futuristic approach to a lot of these things. and if his trends indeed 463 00:27:42,910 --> 00:27:46,570 Ejaaz: end up playing out like he early backed Nvidia and Cerebrus, 464 00:27:46,800 --> 00:27:49,880 Ejaaz: you could end up having like exponential gains over the next couple of years. 465 00:27:50,040 --> 00:27:53,700 Ejaaz: But again, it's all based on his one thesis, which is we are not in a bubble. 466 00:27:54,020 --> 00:27:56,800 Ejaaz: I'm curious whether the listeners of this show agree with this. 467 00:27:57,700 --> 00:28:00,680 Ejaaz: Obviously, most people aren't as technical or in the weeds as Gavin, 468 00:28:00,900 --> 00:28:03,060 Ejaaz: but just generally speaking, after you've heard this episode, 469 00:28:03,420 --> 00:28:04,620 Ejaaz: do you think we're in a bubble? 470 00:28:04,800 --> 00:28:07,420 Ejaaz: Do you think we're not? What are the arguments for and against? 471 00:28:07,520 --> 00:28:08,880 Ejaaz: Is there anything that we particularly missed? 472 00:28:09,380 --> 00:28:11,900 Ejaaz: I don't know, Josh, do you think we're in a bubble as we round up? 473 00:28:12,240 --> 00:28:15,200 Josh: I think we're certainly in a bubble. Where we are in that bubble is to be debated. 474 00:28:15,200 --> 00:28:17,040 Josh: It seems like it's in the earlier stages. 475 00:28:17,260 --> 00:28:19,500 Josh: So hopefully it continues to be that way. According to Gavin, 476 00:28:19,780 --> 00:28:24,740 Josh: so long as TSMC continues to limit their ability to produce chips, it will be fine. 477 00:28:24,940 --> 00:28:27,400 Josh: But that's the outlook. We have Leopold now that we've covered, 478 00:28:27,660 --> 00:28:29,740 Josh: whose success is measured in quarters. 479 00:28:30,000 --> 00:28:32,400 Josh: We have Gavin, whose success is measured in decades. 480 00:28:32,820 --> 00:28:35,280 Josh: And perhaps somewhere in the middle is where a lot of people think of themselves. 481 00:28:35,680 --> 00:28:38,400 Josh: So if you did enjoy, please don't forget to share this with your friend. 482 00:28:38,660 --> 00:28:42,320 Josh: Let us know which ones you are most. If it's not a thesis, perhaps there's a 483 00:28:42,320 --> 00:28:46,440 Josh: ticker that we should be looking at. I think it's exciting time because everything's moving quickly. 484 00:28:46,540 --> 00:28:49,040 Josh: Whether it be up or down, there is a lot of movement. There's a lot of volatility. 485 00:28:49,360 --> 00:28:50,540 Josh: It's fun to get involved in. 486 00:28:50,700 --> 00:28:51,320 Ejaaz: See you guys tomorrow. 487 00:28:51,820 --> 00:28:52,180 Josh: See you tomorrow.