1 00:00:00,020 --> 00:00:05,800 Ejaaz: The most contrarian AI bet of 2026 is the same company that you buy toilet paper from. 2 00:00:06,020 --> 00:00:09,280 Ejaaz: Many people will mistake Amazon as just an e-commerce company, 3 00:00:09,400 --> 00:00:11,440 Ejaaz: but they're secretly a frontier AI lab. 4 00:00:11,600 --> 00:00:15,000 Ejaaz: For example, did you know that they manufacture their own AI chips that are 5 00:00:15,000 --> 00:00:19,880 Ejaaz: as good as NVIDIA's GPUs, but 50% cheaper, which means that companies like Anthropic 6 00:00:19,880 --> 00:00:24,020 Ejaaz: and OpenAI, which have signed deals with Amazon, save tens of billions of dollars 7 00:00:24,020 --> 00:00:26,360 Ejaaz: training their frontier models. But that's not all. 8 00:00:26,680 --> 00:00:30,940 Ejaaz: Amazon's compute platform, which accounts for 50% of their operating profits, AWS. 9 00:00:31,380 --> 00:00:35,900 Ejaaz: They're running the same playbook for AI now, serving AI cloud to any company 10 00:00:35,900 --> 00:00:38,020 Ejaaz: that wants to inference or train their own models. 11 00:00:38,320 --> 00:00:41,740 Ejaaz: And finally, Amazon has a secret up their sleeve which no one's talking about. 12 00:00:42,100 --> 00:00:45,760 Ejaaz: Robots. For over a decade, Amazon robots have helped them scale manufacturing 13 00:00:45,760 --> 00:00:49,120 Ejaaz: and factory automation to the tune of tens of billions of dollars, 14 00:00:49,260 --> 00:00:53,520 Ejaaz: which make them the perfect company to design and build the robots of tomorrow. 15 00:00:53,920 --> 00:00:58,960 Ejaaz: Amazon is easily a $5 trillion company hidden as a shopping platform. 16 00:00:59,200 --> 00:01:02,260 Josh: It's funny you mentioned the shopping platform part because when everyone, 17 00:01:02,260 --> 00:01:05,820 Josh: myself included, thinks of Amazon, they very clearly think of shopping platform. 18 00:01:06,000 --> 00:01:08,980 Josh: And for the right reason, I have some pretty unbelievable stats. 19 00:01:09,360 --> 00:01:14,680 Josh: So last year, Amazon shipped 6.3 billion packages, which is 17 and a quarter 20 00:01:14,680 --> 00:01:18,520 Josh: million per day, which is 200 per second that gets shipped. 21 00:01:18,740 --> 00:01:22,920 Josh: So that equates to about 30% of all the US parcels were one company. 22 00:01:23,080 --> 00:01:25,760 Josh: And that means one out of every three and a half packages was shipped by amazon 23 00:01:25,760 --> 00:01:27,080 Josh: so in terms of the amount of 24 00:01:27,520 --> 00:01:32,040 Josh: mass and atoms that they're moving throughout the universe throughout the world 25 00:01:32,040 --> 00:01:35,200 Josh: like as it relates to the other mag 7 companies they're moving the most amount 26 00:01:35,200 --> 00:01:38,860 Josh: of stuff just raw stuff and you have to imagine that once you start to apply 27 00:01:38,860 --> 00:01:42,800 Josh: ai to this in terms of efficiency gains and improvements like you're mentioning with robotics, 28 00:01:43,300 --> 00:01:47,900 Josh: The amount of stuff can really be optimized quite a bit and have a meaningful impact on the business. 29 00:01:48,040 --> 00:01:50,760 Josh: But what we're seeing with the stock price that's on the screen here tells a 30 00:01:50,760 --> 00:01:55,320 Josh: very different story, which is basically flat in a year where every company 31 00:01:55,320 --> 00:02:00,340 Josh: in the world that was building an AI at Amazon's size went up an outrageous amount. 32 00:02:00,640 --> 00:02:04,620 Ejaaz: I think the story behind this is Amazon's just very misunderstood. 33 00:02:04,940 --> 00:02:10,580 Ejaaz: Like, to your point, in the last year, it's gone up 1.4%, which is just insane. 34 00:02:10,620 --> 00:02:13,860 Ejaaz: It doesn't even beat inflation. doesn't beat inflation, which is just insane. 35 00:02:14,040 --> 00:02:17,800 Ejaaz: And listen, there are theories as to why this might be, but we're here to tell 36 00:02:17,800 --> 00:02:22,720 Ejaaz: you the story why Amazon is basically the biggest AI beach ball underwater that 37 00:02:22,720 --> 00:02:24,560 Ejaaz: is about to pop up in 2026. 38 00:02:24,900 --> 00:02:29,580 Ejaaz: And some of my foundational thesis behind this, Josh, is that the stuff that 39 00:02:29,580 --> 00:02:31,880 Ejaaz: they focus on is really unsexy. 40 00:02:32,040 --> 00:02:35,160 Ejaaz: It's operational. And they're about to do the same for AI. 41 00:02:35,320 --> 00:02:38,920 Ejaaz: Like, think about it, like sorting, fulfilling packages, delivering it to people 42 00:02:38,920 --> 00:02:40,620 Ejaaz: isn't really a sexy thing. 43 00:02:40,800 --> 00:02:44,080 Ejaaz: And then if you talk about compute and serving compute to different companies, 44 00:02:44,280 --> 00:02:46,300 Ejaaz: again, I don't really care about it. But 45 00:02:46,700 --> 00:02:51,760 Ejaaz: What most people don't realize is that the top software companies in the world run on AWS. 46 00:02:52,060 --> 00:02:56,240 Ejaaz: That's why when AWS servers went out a few weeks ago, the world couldn't function. 47 00:02:56,360 --> 00:02:59,120 Ejaaz: I couldn't use X. I couldn't scroll my favorite social media platforms, 48 00:02:59,240 --> 00:03:01,120 Ejaaz: Josh, because it ran on AWS. 49 00:03:01,260 --> 00:03:04,320 Ejaaz: So most people don't realize this. And they're about to do the same for AI. 50 00:03:04,660 --> 00:03:06,860 Josh: Yeah. And one last thing I want to mention on this chart, since we have it on 51 00:03:06,860 --> 00:03:10,140 Josh: the screen, is that little PE ratio number, the price to earnings ratio, 52 00:03:10,320 --> 00:03:15,540 Josh: it's down to 35 now. And for reference, back in 2018, Amazon was trading at 53 00:03:15,540 --> 00:03:18,240 Josh: over a 200 times price to earnings ratio. 54 00:03:18,440 --> 00:03:22,380 Josh: So the multiples have really compressed a lot. But this is a much more mature 55 00:03:22,380 --> 00:03:27,280 Josh: company, which actually has a vector of growth, which is adding AI to the mix 56 00:03:27,280 --> 00:03:28,180 Josh: to increase this efficiency. 57 00:03:28,440 --> 00:03:31,740 Josh: So EJS, you wrote all about this. It's in the newsletter that we published. 58 00:03:31,880 --> 00:03:35,960 Josh: But you actually created a proper article going through the bull case for Amazon. 59 00:03:35,960 --> 00:03:38,720 Josh: And I think we're going to spend a good amount of time kind of going through 60 00:03:38,720 --> 00:03:42,440 Josh: the outline that you framed here for why you believe it to be, 61 00:03:42,720 --> 00:03:45,920 Josh: I mean, as the title says, the most asymmetric bet of 2026. 62 00:03:46,360 --> 00:03:49,100 Ejaaz: And for those of you who are wondering, hey, what is this article? 63 00:03:49,180 --> 00:03:51,120 Ejaaz: What is this newsletter? It's the Limitless Newsletter. 64 00:03:51,320 --> 00:03:55,060 Ejaaz: And if you were subscribed to it, you would have seen this article about a week and a half ago. 65 00:03:55,220 --> 00:04:00,540 Ejaaz: So again, if you want the best alpha in AI, you have to subscribe to our newsletter. 66 00:04:00,680 --> 00:04:04,420 Ejaaz: But yes, Josh, one of the first things or one of my first arguments as to why 67 00:04:04,420 --> 00:04:09,080 Ejaaz: Amazon and there's an asymmetric bet here, is this thing called AI chips. 68 00:04:09,160 --> 00:04:11,800 Ejaaz: You heard of them, Josh? You know, these little GPUs thing? 69 00:04:12,020 --> 00:04:16,160 Josh: You come across them? I might have heard of GPU, a little TPU, a whole bunch of U's. 70 00:04:16,380 --> 00:04:19,520 Ejaaz: All right, all right, okay. So you've heard of NVIDIA, it sounds like. 71 00:04:19,640 --> 00:04:23,000 Ejaaz: It sounds like you've heard of Google as well, as you know we're very bullish on Google here. 72 00:04:23,420 --> 00:04:27,740 Ejaaz: But what most people don't realize is Amazon created their own chips. 73 00:04:27,900 --> 00:04:29,440 Ejaaz: And it wasn't like they did this yesterday. 74 00:04:29,660 --> 00:04:35,400 Ejaaz: They did this 10 years ago, Josh, when they acquired a company called Annapurna Labs in 2015, 75 00:04:35,660 --> 00:04:41,580 Ejaaz: which marks the start of their training and build of AI chips to help them with 76 00:04:41,580 --> 00:04:44,960 Ejaaz: machine learning inference, stuff like they was figuring out way back when, 77 00:04:45,020 --> 00:04:48,140 Ejaaz: when it came to recommend systems on their shopping platform. 78 00:04:48,400 --> 00:04:51,560 Ejaaz: Now, if you fast forward today, in the last few weeks, Josh, 79 00:04:51,660 --> 00:04:55,540 Ejaaz: they released this chip called Tranium 3. It's part of their Tranium series 80 00:04:55,540 --> 00:05:01,800 Ejaaz: of chips, which are used to build and scale large LLMs or rather large AI models. 81 00:05:02,260 --> 00:05:05,780 Ejaaz: And Tranium 3, obviously, you know, the first question that pops into your head 82 00:05:05,780 --> 00:05:07,660 Ejaaz: is like, well, how does this compare to NVIDIA? 83 00:05:07,880 --> 00:05:10,800 Ejaaz: Well, let me give you a few stats to kind of whet your appetite, 84 00:05:10,960 --> 00:05:15,260 Ejaaz: Josh, which is it is four and a half times more powerful than Tranium 2, 85 00:05:15,460 --> 00:05:18,580 Ejaaz: but it's also four times cheaper than Tranium 2. 86 00:05:18,720 --> 00:05:22,700 Ejaaz: So if you net both of those together, you get like a 10X kind of chip here, right? 87 00:05:23,200 --> 00:05:28,220 Ejaaz: But then it can also store five times more data than the previous chip that they had. 88 00:05:28,360 --> 00:05:33,880 Ejaaz: So all of these combined together gives you about 80 to 90% of the same performance 89 00:05:33,880 --> 00:05:37,040 Ejaaz: as NVIDIA's latest GPU, Blackwell. 90 00:05:37,340 --> 00:05:40,420 Ejaaz: So I'm not just talking about NVIDIA's second, third, or fourth generation. 91 00:05:40,720 --> 00:05:44,740 Ejaaz: I'm talking about the latest generation that they have right now. So automatically, 92 00:05:45,020 --> 00:05:48,840 Ejaaz: Josh, if I was a Frontier AI lab that is spending hundreds of billions of dollars 93 00:05:48,840 --> 00:05:55,800 Ejaaz: each year on NVIDIA's GPUs and you suddenly give me an option to spend 50% of that bill. So I save... 94 00:05:56,360 --> 00:05:59,360 Ejaaz: 50 billion dollars why wouldn't i use this chip 95 00:05:59,360 --> 00:06:02,000 Josh: Yeah it sounds like a pretty good deal and i think we should 96 00:06:02,000 --> 00:06:05,120 Josh: probably start by outlining what exactly this chip is because there's a difference between 97 00:06:05,120 --> 00:06:11,240 Josh: nvidia's gpu google's tpu and then training which is something totally different 98 00:06:11,240 --> 00:06:15,660 Josh: and they fit if i'm not mistaken into two kind of basic categories so gpus which 99 00:06:15,660 --> 00:06:19,720 Josh: is nvidia's blackwell system that is a graphical processing unit it's kind of 100 00:06:19,720 --> 00:06:23,360 Josh: a general purpose supercomputer you can use it for a lot of different things, 101 00:06:23,480 --> 00:06:24,300 Josh: graphic being one of them, 102 00:06:24,480 --> 00:06:28,440 Josh: but also the matrix math required to do AI training as another. 103 00:06:28,980 --> 00:06:32,180 Josh: With TPUs like Google's making and these Amazon chips through Tranium, 104 00:06:32,460 --> 00:06:35,420 Josh: they're more focused on specific types of math. 105 00:06:35,580 --> 00:06:39,860 Josh: They are not general purpose. They're narrowly focused on AI training and AI 106 00:06:39,860 --> 00:06:42,820 Josh: inference. And that's where you get a lot of the efficiency improvements. 107 00:06:42,960 --> 00:06:45,540 Josh: So like we're seeing on the screen, 4.4 times higher improvement, 108 00:06:45,760 --> 00:06:48,200 Josh: four times more memory bandwidth, like the specs are insane, 109 00:06:48,460 --> 00:06:50,440 Josh: but they cannot be used for everything. 110 00:06:50,580 --> 00:06:53,660 Josh: So it's a very specific type of customer that wants these types of chips. 111 00:06:53,880 --> 00:06:59,360 Ejaaz: Exactly. So if you were to kind of like compare the two NVIDIA GPUs and Amazon's training chips, 112 00:06:59,720 --> 00:07:04,560 Ejaaz: NVIDIA's GPUs can be used for a lot of broad use cases when it comes to training models. 113 00:07:04,720 --> 00:07:07,080 Ejaaz: Like there are several different types of ways to train an AI model. 114 00:07:07,240 --> 00:07:11,020 Ejaaz: If you weren't sure which one to use, you would probably use an NVIDIA GPU because 115 00:07:11,020 --> 00:07:13,120 Ejaaz: it would just be consistent across all of those things. 116 00:07:13,400 --> 00:07:17,940 Ejaaz: But for Amazon's training chips, you need to specifically know exactly how you're 117 00:07:17,940 --> 00:07:21,040 Ejaaz: going to train a specific model and then it ends up being cheaper. 118 00:07:21,160 --> 00:07:25,900 Ejaaz: So it's for like a highly specialized type of AI lab that wants to train their own model. 119 00:07:26,080 --> 00:07:31,420 Ejaaz: And what's interesting about this, Josh, is it's not just kind of the specific 120 00:07:31,420 --> 00:07:35,760 Ejaaz: architecture of how this chip is designed, it's also very much the cost. 121 00:07:35,840 --> 00:07:40,080 Ejaaz: I have a table pulled up here, and if you can take a direct look at it, 122 00:07:40,180 --> 00:07:44,980 Ejaaz: which compares Tranium 3 to Google's TPUs and NVIDIA's latest chip, 123 00:07:45,120 --> 00:07:49,480 Ejaaz: the Blackwell, and it is half the cost of NVIDIA's Blackwell. 124 00:07:49,760 --> 00:07:53,860 Ejaaz: That is excluding NVIDIA's margin that they add on top of this, Josh. 125 00:07:53,980 --> 00:07:58,360 Ejaaz: So if you look at this, the average cost of an NVIDIA chip to manufacture is 126 00:07:58,360 --> 00:08:03,260 Ejaaz: around $6,500 to $7,000, but they end up selling it for $40,000. 127 00:08:03,720 --> 00:08:07,120 Ejaaz: That's the price that Amazon has to pay to buy these chips. That's the cost 128 00:08:07,120 --> 00:08:09,380 Ejaaz: that OpenAI has to pay to buy these chips. 129 00:08:09,660 --> 00:08:14,060 Ejaaz: So I can think of two things here. If Amazon has an almost as good chip, 130 00:08:14,420 --> 00:08:18,260 Ejaaz: then it's going to largely kind of be a more attractive chip to buy for Frontier 131 00:08:18,260 --> 00:08:22,020 Ejaaz: AI Labs, but it's also going to cut into NVIDIA's margins drastically. 132 00:08:22,020 --> 00:08:26,360 Ejaaz: So you start seeing Amazon and Google TPUs being able to eat into the market 133 00:08:26,360 --> 00:08:27,820 Ejaaz: monopoly that NVIDIA has. 134 00:08:28,240 --> 00:08:30,480 Josh: Okay, so interesting. I want you to kind of help me understand this, 135 00:08:30,540 --> 00:08:34,940 Josh: because a lot of this has been new information to me. I got intro to this through the article, and, 136 00:08:35,450 --> 00:08:42,190 Josh: What I understand is that Amazon's chips, not TPUs, are better on a per watt 137 00:08:42,190 --> 00:08:44,230 Josh: basis. They're more efficient. They're more cost effective. 138 00:08:44,450 --> 00:08:48,430 Josh: But I guess my question is, if Amazon were to start making these available to 139 00:08:48,430 --> 00:08:51,470 Josh: the public tomorrow, would they outsell Blackwell? 140 00:08:51,550 --> 00:08:54,990 Josh: Would people be more interested in these or would it just be a very specific audience? 141 00:08:55,310 --> 00:08:59,150 Ejaaz: No. OK, so there's two things I want to bring up with you that kind of like 142 00:08:59,150 --> 00:09:01,770 Ejaaz: sway people's decisions when they're choosing between the chips. 143 00:09:02,370 --> 00:09:07,430 Ejaaz: Number one, an NVIDIA Blackwell chip on its own and also an Amazon training 144 00:09:07,430 --> 00:09:09,210 Ejaaz: chip on its own isn't useful, Josh. 145 00:09:09,370 --> 00:09:12,190 Ejaaz: You need to stack them into this thing called clusters. 146 00:09:12,810 --> 00:09:17,550 Ejaaz: Now, for an NVIDIA chip, you only have to put 72 of these chips together to 147 00:09:17,550 --> 00:09:19,470 Ejaaz: get the same performance per watt. 148 00:09:19,650 --> 00:09:23,710 Ejaaz: But with an Amazon chip, you need to stack 144 of them together. 149 00:09:23,710 --> 00:09:25,450 Ejaaz: So it's a higher volume thing. 150 00:09:25,650 --> 00:09:29,310 Josh: Okay, so it's about double the amount of chips per little cluster that we have 151 00:09:29,310 --> 00:09:34,530 Josh: in a rack. and do those ships are they more expensive too or are they like cost 152 00:09:34,530 --> 00:09:38,630 Josh: better because it seems like that's a lot more complexity for like 80% of the efficiency. 153 00:09:39,530 --> 00:09:42,630 Ejaaz: They're cheaper. So on this table right now, they're half the cost. 154 00:09:42,750 --> 00:09:46,170 Ejaaz: That's where the $3,000, the $3,500 come from. 155 00:09:46,330 --> 00:09:50,290 Ejaaz: So you may have more chips, which may take up more volume in your data center, 156 00:09:50,290 --> 00:09:54,410 Ejaaz: but they are cheaper to run at a cost kind of inference, right? 157 00:09:54,810 --> 00:09:58,610 Ejaaz: The other thing, Josh, is you might have noticed I said they're not as good 158 00:09:58,610 --> 00:10:02,170 Ejaaz: as NVIDIA GPUs. They're around 80% to 90% as good. 159 00:10:02,350 --> 00:10:07,670 Ejaaz: And the reason why there's that difference is because of NVIDIA's software moat. 160 00:10:07,730 --> 00:10:12,090 Ejaaz: So this is something called CUDA or Compute Unified Device Architecture. 161 00:10:12,510 --> 00:10:16,050 Ejaaz: So basically, if you have the chips, that doesn't solve your entire problem. 162 00:10:16,170 --> 00:10:20,510 Ejaaz: You need software to be able to make these chips run really coherently together in their clusters. 163 00:10:21,110 --> 00:10:26,350 Ejaaz: NVIDIA has the stronghold of this, Josh. Any AI lab that is using NVIDIA GPUs, 164 00:10:26,430 --> 00:10:30,630 Ejaaz: which is the majority of AI labs, run the CUDA software system. 165 00:10:30,810 --> 00:10:34,390 Ejaaz: And typically, this has locked customers into using NVIDIA. 166 00:10:34,530 --> 00:10:39,050 Ejaaz: So let's say they're interested in using Google's TPUs, they may not necessarily 167 00:10:39,050 --> 00:10:42,870 Ejaaz: still want to jump to use Google's TPUs because the software isn't the same. 168 00:10:43,030 --> 00:10:45,990 Ejaaz: They would have to rewrite their entire code base. 169 00:10:46,350 --> 00:10:50,110 Ejaaz: Amazon saw that and thought, hmm, I bet I could make this easier. 170 00:10:50,310 --> 00:10:54,310 Ejaaz: And so they released this thing called a Neuron SDK, which now allows you to 171 00:10:54,310 --> 00:10:59,430 Ejaaz: copy and paste your code base from your NVIDIA instance into your Amazon GPU 172 00:10:59,430 --> 00:11:02,630 Ejaaz: or your Amazon Tranium chip in a few clicks. 173 00:11:02,870 --> 00:11:05,990 Josh: Maybe this is a good time to point out the fact that 174 00:11:05,990 --> 00:11:08,750 Josh: this is a really big deal for amazon even if they don't sell 175 00:11:08,750 --> 00:11:13,130 Josh: these chips anywhere um because aws is such a huge story we were looking up 176 00:11:13,130 --> 00:11:17,290 Josh: before this episode was recorded how much of the internet is run by aws and 177 00:11:17,290 --> 00:11:21,350 Josh: it's about a third and another fascinating thing that you kind of pointed out 178 00:11:21,350 --> 00:11:26,490 Josh: in the intro but um as of last quarter i believe um aws 179 00:11:26,890 --> 00:11:31,250 Josh: revenue was something like less than 20% of the total company, 180 00:11:31,530 --> 00:11:36,410 Josh: but it accounts for close to 70% of the actual profit share. 181 00:11:36,590 --> 00:11:38,530 Josh: I think it's like 66% as of the third quarter. 182 00:11:38,730 --> 00:11:43,570 Josh: So this small part of Amazon's business accounts for over half of the total 183 00:11:43,570 --> 00:11:45,270 Josh: profit every single quarter that comes in. 184 00:11:45,370 --> 00:11:49,610 Josh: And with these plans to scale by using these new shifts, by using this gigantic 185 00:11:49,610 --> 00:11:51,810 Josh: data center that we're seeing on the screen that we're going to get into, 186 00:11:52,070 --> 00:11:55,590 Josh: there is a very clear trajectory for amplifying the one part of the business 187 00:11:55,590 --> 00:11:57,910 Josh: that actually matters the most for their top line. 188 00:11:58,050 --> 00:12:01,750 Josh: And we were talking earlier, it's kind of similar to what Costco does with their 189 00:12:01,750 --> 00:12:05,910 Josh: membership, where Costco's business operates on very thin profit margins. 190 00:12:05,970 --> 00:12:09,330 Josh: They really don't make a lot, if any money, on the actual goods sold. 191 00:12:09,570 --> 00:12:12,750 Josh: A lot of that revenue comes from that membership, from the lock-in. 192 00:12:12,870 --> 00:12:17,910 Josh: And Amazon, what they have with AWS, is this unbelievably profitable engine 193 00:12:17,910 --> 00:12:20,770 Josh: that is becoming much more efficient using these new Tranium chips. 194 00:12:20,890 --> 00:12:23,470 Josh: But here on screen, and we have this really crazy looking data center. 195 00:12:23,570 --> 00:12:25,750 Josh: So maybe you could tell us more about what's going on here. 196 00:12:26,390 --> 00:12:30,650 Ejaaz: Yeah, I mean, listen, we're not strangers to crazy data center setups. 197 00:12:30,730 --> 00:12:36,010 Ejaaz: We've spoken about Elon Musk's Colossus 2. We've spoken about Meta's super data 198 00:12:36,010 --> 00:12:36,650 Ejaaz: center that they're building. 199 00:12:36,750 --> 00:12:39,590 Ejaaz: Basically, all the top companies are spending tens of billions of dollars. 200 00:12:39,810 --> 00:12:43,330 Ejaaz: And Amazon is no stranger to this either. They've invested $11 billion and they're 201 00:12:43,330 --> 00:12:47,150 Ejaaz: going to invest another $20 billion next year to build out their Indiana data 202 00:12:47,150 --> 00:12:51,770 Ejaaz: center campus to create around 2.2 gigawatts of compute. 203 00:12:51,910 --> 00:12:56,690 Ejaaz: That's equitable to about 1 million homes worth of energy by the end of next 204 00:12:56,690 --> 00:12:59,570 Ejaaz: year, which is just an insane goal to kind of like figure out. 205 00:12:59,890 --> 00:13:03,870 Ejaaz: And to your point, Josh, like if they're able to pull off what they did with 206 00:13:03,870 --> 00:13:06,550 Ejaaz: AWS for AI compute specifically, 207 00:13:07,260 --> 00:13:11,640 Ejaaz: There is kind of like no reason for me to think at least why they wouldn't eat 208 00:13:11,640 --> 00:13:15,600 Ejaaz: into all the other NeoCloud's valuations. We've spoken about, what's his name? 209 00:13:15,940 --> 00:13:20,260 Ejaaz: Leopold investing in this company called CoreWeave, which was one of the top NeoCloud providers. 210 00:13:20,500 --> 00:13:24,240 Ejaaz: And he invested to the tune of like, I think it was like 350 billion, 211 00:13:24,340 --> 00:13:26,620 Ejaaz: almost $400 billion into this company. 212 00:13:26,980 --> 00:13:30,420 Ejaaz: If Amazon just eats, like switches this on at the end of next year, 213 00:13:30,620 --> 00:13:35,380 Ejaaz: they like companies who are already running on AWS will just switch to the AI 214 00:13:35,380 --> 00:13:39,240 Ejaaz: version of AWS. it kind of doesn't make sense for them to kind of flip here. 215 00:13:39,500 --> 00:13:45,200 Ejaaz: And it's why I think they have such a kind of sticky mode. They help with all the unsexy stuff, Josh. 216 00:13:45,480 --> 00:13:49,380 Ejaaz: I don't know if you saw this rollout of something called AI Factory. 217 00:13:49,540 --> 00:13:50,860 Ejaaz: Did you catch this by any chance? 218 00:13:51,320 --> 00:13:54,360 Josh: No. AI Factory sounds interesting. We like factories. Okay. 219 00:13:54,700 --> 00:13:59,920 Ejaaz: So let me lay this out for you. A lot of enterprises and even governments want 220 00:13:59,920 --> 00:14:04,460 Ejaaz: to create their own versions of AI models, but they run into two main issues. 221 00:14:04,820 --> 00:14:08,340 Ejaaz: Number one, they don't know who to buy the chips from. Should they go to NVIDIA? 222 00:14:08,420 --> 00:14:10,700 Ejaaz: Should they go to Google? Should they go to Amazon? I have no idea. 223 00:14:11,160 --> 00:14:14,340 Ejaaz: And then two, they don't want to set it up themselves, right? 224 00:14:14,580 --> 00:14:19,460 Ejaaz: But they also don't want to rent compute directly through AWS. Why? 225 00:14:19,800 --> 00:14:23,100 Ejaaz: Because, you know, they have some private information. They don't want to leak information. 226 00:14:23,260 --> 00:14:26,060 Ejaaz: They want Amazon to own the information. They want to hold it on their own private 227 00:14:26,060 --> 00:14:28,200 Ejaaz: service. So Amazon looked at this and said, 228 00:14:28,600 --> 00:14:32,640 Ejaaz: Okay, we're rolling out a service called AWS Factory, and here's what we offer. 229 00:14:32,800 --> 00:14:38,120 Ejaaz: If you want NVIDIA GPUs, we got you. If you want Amazon Tranium GPUs or chips, 230 00:14:38,360 --> 00:14:40,860 Ejaaz: which are, by the way, 50% cheaper, we also got you. 231 00:14:40,960 --> 00:14:43,840 Ejaaz: If you want a hybrid or mix of both of these things, we've got you. 232 00:14:43,920 --> 00:14:47,660 Ejaaz: And what they do is they build the server racks, Josh, for them. 233 00:14:47,740 --> 00:14:51,840 Ejaaz: They basically build out a data center for them, and this benefits them in so 234 00:14:51,840 --> 00:14:54,680 Ejaaz: many ways because they have the software and they have the hardware, 235 00:14:54,680 --> 00:14:59,440 Ejaaz: and they have 50% less the cost if they run AWS or Amazon Tranium chips. 236 00:14:59,600 --> 00:15:02,360 Ejaaz: So it's like an all-in-one package where they have lower latency, 237 00:15:02,360 --> 00:15:06,260 Ejaaz: they save on costs, and they have premium frontier intelligence. Pretty cool. 238 00:15:06,340 --> 00:15:07,720 Josh: But there's an important distinction 239 00:15:07,720 --> 00:15:10,980 Josh: there where they're not building the factories for the company. 240 00:15:11,140 --> 00:15:14,780 Josh: They're building their own infrastructure that they're lending to the company, right? 241 00:15:14,920 --> 00:15:19,400 Josh: So if we're considering a company like Oracle, whose job is to build data centers 242 00:15:19,400 --> 00:15:22,680 Josh: for companies, you can think about the Project Stargate project with OpenAI, 243 00:15:23,120 --> 00:15:26,320 Josh: they are building actual infrastructure that OpenAI will own. 244 00:15:26,660 --> 00:15:29,960 Josh: What Amazon is doing is they are building the infrastructure for you, 245 00:15:30,080 --> 00:15:34,620 Josh: but they're just leasing it to you. They still own the underlying core infra. 246 00:15:35,080 --> 00:15:38,540 Josh: So these other companies are essentially bankrolling the build-out of these 247 00:15:38,540 --> 00:15:41,900 Josh: factories, but doing so in a way that doesn't incur a lot of debt. 248 00:15:42,000 --> 00:15:43,500 Josh: Like they already have customers here. 249 00:15:43,640 --> 00:15:47,540 Josh: And the switching costs, I was looking because I was curious as it relates to 250 00:15:47,540 --> 00:15:50,700 Josh: these big cloud providers. Like AWS is 30%. 251 00:15:51,200 --> 00:15:55,480 Josh: Azure and Google Cloud are another 33%. And then the rest is just kind of this 252 00:15:55,480 --> 00:16:00,040 Josh: mix of things. So a third of the companies are already, they trust the security 253 00:16:00,040 --> 00:16:02,160 Josh: of Amazon. They already use Amazon. 254 00:16:02,260 --> 00:16:04,500 Josh: They have their whole custom software stack built on Amazon. 255 00:16:04,680 --> 00:16:09,180 Josh: And in the world that they can just extend this to integrate AI into their offering, 256 00:16:09,300 --> 00:16:13,100 Josh: well, now 33% of the internet, they just get to direct AI offering built in. 257 00:16:13,180 --> 00:16:14,680 Josh: And that's like a pretty powerful thing. 258 00:16:15,100 --> 00:16:19,660 Ejaaz: Yeah, Josh, you know which other little small company this reminds me of? 259 00:16:19,840 --> 00:16:20,860 Josh: Google. Who's that? 260 00:16:21,060 --> 00:16:25,520 Ejaaz: Google already had their roots sewn into so many different products. 261 00:16:25,700 --> 00:16:29,360 Ejaaz: Basically, anyone who's ever graced the internet has come across a Google product, 262 00:16:29,500 --> 00:16:34,160 Ejaaz: whether it's Gmail, Android, Play Store, whether it's Google Search itself. 263 00:16:34,620 --> 00:16:39,480 Ejaaz: Amazon is the same bedrock for this, except it's just, hey, we've got the compute 264 00:16:39,480 --> 00:16:43,220 Ejaaz: that you're going to run your number one website or internet product on. 265 00:16:43,220 --> 00:16:47,460 Ejaaz: And they're converting the same users, as you said, into AI users. 266 00:16:47,580 --> 00:16:49,580 Ejaaz: I just think it's a complete no-brainer. 267 00:16:49,760 --> 00:16:54,080 Ejaaz: And yeah, to kind of build on your analogy, they're not building the entire data center for you. 268 00:16:54,160 --> 00:16:57,180 Ejaaz: You're going to still have to buy the warehouse and supply it with energy and 269 00:16:57,180 --> 00:16:59,700 Ejaaz: electrical grid, but they've got everything else for you. 270 00:16:59,780 --> 00:17:03,040 Ejaaz: You don't have to have the upfront AI capex costs that, for example, 271 00:17:03,180 --> 00:17:08,340 Ejaaz: OpenAI has when they're committing $1.4 trillion over the next five years to buy these GPUs. 272 00:17:08,480 --> 00:17:11,920 Josh: Okay, so we've covered the chips, we've covered the cloud, Now we have the physical 273 00:17:11,920 --> 00:17:16,760 Josh: infrastructure, the world of atoms. This gets to our 6.3 billion package statistic 274 00:17:16,760 --> 00:17:22,260 Josh: where Amazon, of all the Mag7 companies, they just move the most amount of stuff through the universe. 275 00:17:22,480 --> 00:17:26,360 Josh: And through that, they benefit, they stand to benefit a lot from automation, 276 00:17:26,560 --> 00:17:28,380 Josh: particularly as it relates to robots. 277 00:17:28,520 --> 00:17:32,260 Josh: So here we have news that they have three quarters of a million robots already deployed. 278 00:17:32,400 --> 00:17:35,580 Josh: EJS, what are their plans going forward that are part of your bull case as it 279 00:17:35,580 --> 00:17:37,560 Josh: relates to kind of robotics and warehouses? 280 00:17:37,880 --> 00:17:40,160 Ejaaz: Okay, what I'm about to say is going to sound controversial. 281 00:17:40,880 --> 00:17:47,200 Ejaaz: Because I have said that Tesla is the number one robot company so many times, but I was wrong. 282 00:17:47,600 --> 00:17:50,780 Ejaaz: It's going to be- Oh, wow, that is a hot take. It's going to be Amazon. 283 00:17:50,780 --> 00:17:54,480 Ejaaz: They already have almost a million of these things out there. And listen, listen. 284 00:17:54,680 --> 00:17:58,660 Ejaaz: Okay. They might not be the humanoid robots that grace your home, 285 00:17:58,880 --> 00:17:59,940 Ejaaz: that help you with the laundry. 286 00:18:00,100 --> 00:18:03,260 Ejaaz: I have laundry in my thing that needs to get sorted right now. 287 00:18:03,320 --> 00:18:04,420 Ejaaz: Wish I had one of those, right? 288 00:18:04,680 --> 00:18:09,440 Ejaaz: It's not going to be a robot car that takes me wherever I need to from A to 289 00:18:09,440 --> 00:18:12,660 Ejaaz: B with minimal accidental type stuff. 290 00:18:13,180 --> 00:18:16,720 Ejaaz: But they're going to be the robots that scale very important things, 291 00:18:16,880 --> 00:18:20,820 Ejaaz: such as manual labor, Josh, which is like, you know, a multi-trillion dollar 292 00:18:20,820 --> 00:18:23,020 Ejaaz: industry or global sector as itself. 293 00:18:23,060 --> 00:18:28,440 Ejaaz: If you are able to cut down costs by 50 to maybe even additional percent, 294 00:18:28,840 --> 00:18:30,100 Ejaaz: why wouldn't you do that? 295 00:18:30,300 --> 00:18:34,020 Ejaaz: Amazon is the perfectly positioned company for that. They already have a million of these robots. 296 00:18:34,340 --> 00:18:38,480 Ejaaz: They are making very, very aggressive cuts on their labor force. 297 00:18:38,620 --> 00:18:42,440 Ejaaz: I don't know if you saw, but like they cut 30,000 jobs about two months ago, 298 00:18:42,620 --> 00:18:47,240 Ejaaz: and they're aiming to scale that up to 600,000 warehouse workers by 2023, 299 00:18:47,240 --> 00:18:51,620 Ejaaz: which is honestly quite scary to hear for a lot of people, I'm sure, who are in these jobs. 300 00:18:51,760 --> 00:18:55,180 Ejaaz: But I think those job roles will essentially evolve. But it basically makes 301 00:18:55,180 --> 00:19:00,560 Ejaaz: Amazon the perfect company to design and build these automation robots of the future. 302 00:19:00,740 --> 00:19:04,300 Ejaaz: Again, they're doing the unsexy stuff, the behind-the-scenes work. 303 00:19:04,440 --> 00:19:08,320 Ejaaz: They're not consumer-facing robots, but it's still a multi-billion-dollar industry. 304 00:19:08,580 --> 00:19:12,740 Josh: We're going to have to agree to disagree on that Tesla robotic statement, I think. 305 00:19:12,820 --> 00:19:16,320 Josh: But I think there is an important difference to outline in the two different 306 00:19:16,320 --> 00:19:17,560 Josh: approaches the companies are taking. 307 00:19:17,800 --> 00:19:23,120 Josh: So Tesla very much treats the factory as the robot, and the robots that they're 308 00:19:23,120 --> 00:19:26,680 Josh: going for are more humanoid general purpose or as it relates to transportation. 309 00:19:26,680 --> 00:19:28,780 Josh: What Amazon is doing and what we're seeing here is... 310 00:19:29,190 --> 00:19:31,890 Josh: Sure, there's some robots that look like humanoids, but a lot of the robots 311 00:19:31,890 --> 00:19:35,210 Josh: that are going to be in these factories, they're narrow-purpose robots. 312 00:19:35,330 --> 00:19:39,790 Josh: They're good for one task. It's a single arm that moves a specific way very quickly. 313 00:19:39,970 --> 00:19:42,710 Josh: And I think that's, when we talk about robots, it's important to understand 314 00:19:42,710 --> 00:19:46,470 Josh: that there are narrow-band robots and general-purpose robots like humanoids. 315 00:19:46,530 --> 00:19:49,490 Josh: And what Amazon most likely stands to benefit from the most, 316 00:19:49,610 --> 00:19:52,970 Josh: at least in the shorter term, is these narrow-band robots like these arms that 317 00:19:52,970 --> 00:19:56,590 Josh: you're seeing on the screen that are really, really efficient at doing one specific 318 00:19:56,590 --> 00:20:00,730 Josh: task. or whether it be those things that are rolling on the floor that can move all these boxes around. 319 00:20:00,970 --> 00:20:05,910 Josh: So as it relates to that type of robotics, Amazon is, they probably have the 320 00:20:05,910 --> 00:20:09,730 Josh: strongest case to be made on how much they can profit from putting those into 321 00:20:09,730 --> 00:20:12,270 Josh: their product because the factory very much is their product. 322 00:20:12,410 --> 00:20:16,890 Josh: How many packages they can ship per minute, if they can get that up from over, what was the number? 323 00:20:16,990 --> 00:20:20,790 Josh: Some crazy number, but whatever 200 per second packages, if they can get that 324 00:20:20,790 --> 00:20:23,590 Josh: up to 250, I mean, that's a huge increase in improvement. 325 00:20:23,750 --> 00:20:26,930 Josh: So I think as it relates to factories, this automation is gonna be huge for them. 326 00:20:27,010 --> 00:20:32,330 Ejaaz: I have a question for you Josh. I noticed you said that it'll help them profit for their own products. 327 00:20:32,490 --> 00:20:37,570 Ejaaz: Do you ever see Amazon selling this type of robot to other factory manufacturers 328 00:20:37,570 --> 00:20:39,210 Ejaaz: in completely different sectors? 329 00:20:39,950 --> 00:20:45,090 Josh: I would hope not. But what I imagine Amazon does is uses this to create more 330 00:20:45,090 --> 00:20:46,930 Josh: of a platform for to entice sellers. 331 00:20:47,110 --> 00:20:51,990 Josh: So a similar business, which I'm kind of thinking, like people use Shopify a 332 00:20:51,990 --> 00:20:53,970 Josh: lot to create web stores and to sell things. 333 00:20:54,390 --> 00:20:57,470 Josh: Shopify is good at creating the tooling. It's good at creating the actual website. 334 00:20:57,650 --> 00:21:01,250 Josh: It's good at pairing customers to consumers, taking care of all the infrastructure. 335 00:21:01,450 --> 00:21:06,970 Josh: But Amazon is the layer that sits beneath that and can actually handle the logistics for you. 336 00:21:07,110 --> 00:21:10,970 Josh: So if you're shipping physical goods, I suspect that Amazon won't sell these 337 00:21:10,970 --> 00:21:12,030 Josh: robots to other companies. 338 00:21:12,170 --> 00:21:14,290 Josh: They will just, again, like they're doing with the data centers, 339 00:21:14,410 --> 00:21:17,930 Josh: they will build the infra and then offer the services to anybody who wants, 340 00:21:18,030 --> 00:21:19,590 Josh: because that's where they get the most amount of profit. 341 00:21:19,670 --> 00:21:24,010 Josh: So if you're a seller on Amazon who wants to know, well, what are the sales 342 00:21:24,010 --> 00:21:26,390 Josh: going to be like in November and December during the holiday season? 343 00:21:26,490 --> 00:21:27,430 Josh: Can you project that for me? 344 00:21:27,590 --> 00:21:31,090 Josh: How much do I need to make? Okay, how much should I send to your warehouse so 345 00:21:31,090 --> 00:21:33,010 Josh: that you're able to get out all the orders on time? 346 00:21:33,070 --> 00:21:37,010 Josh: And as they get more of this intelligence, as they start to build more of an understanding. 347 00:21:37,210 --> 00:21:40,830 Josh: And this gets to the consumer side too, where they understand their customer. 348 00:21:40,950 --> 00:21:43,470 Josh: They know what the customer is shopping for. They know how to sell them ads. 349 00:21:43,650 --> 00:21:47,970 Josh: You get this really fully vertically integrated experience as a seller on Amazon 350 00:21:47,970 --> 00:21:49,750 Josh: that you just can't get anywhere else. 351 00:21:49,830 --> 00:21:52,710 Josh: So if they're building these robots for their own warehouses, 352 00:21:52,970 --> 00:21:56,570 Josh: they should keep them and make everyone else use them because that total vertical 353 00:21:56,570 --> 00:21:58,950 Josh: integration where they understand the customer, they have all the automation, 354 00:21:59,190 --> 00:22:00,790 Josh: it creates the best experience for everybody. 355 00:22:00,910 --> 00:22:02,650 Ejaaz: I see it. I just don't, 356 00:22:03,050 --> 00:22:07,110 Ejaaz: know if I can fully believe it just yet, because like in the same way that they 357 00:22:07,110 --> 00:22:11,710 Ejaaz: rent compute or like kind of CapEx, you know, data hardware to different people, 358 00:22:11,730 --> 00:22:15,150 Ejaaz: I feel like they would probably do the same for their robots as well. 359 00:22:15,730 --> 00:22:20,210 Ejaaz: But I saw Newsleek a few months or like last month, Josh, I don't know if you 360 00:22:20,210 --> 00:22:24,490 Ejaaz: saw this, that they're planning to take on the U.S. Postal Service. 361 00:22:25,570 --> 00:22:28,710 Ejaaz: For those of you who don't know, Amazon pays the U.S. 362 00:22:28,830 --> 00:22:34,710 Ejaaz: Postal Service or rather facilitates $6.6 billion of revenue by using the U.S. Postal Service. 363 00:22:34,930 --> 00:22:38,950 Ejaaz: Amazon just didn't want to scale the delivery kind of service to the extent 364 00:22:38,950 --> 00:22:39,910 Ejaaz: of the U.S. Postal Service. 365 00:22:40,330 --> 00:22:44,210 Ejaaz: Now they're in the market of actually doing that. So we might end up with a 366 00:22:44,210 --> 00:22:47,470 Ejaaz: company that is just all consuming and using all the goods for themselves. 367 00:22:47,730 --> 00:22:51,510 Ejaaz: I don't think they'll do it for chips, but maybe they'll end up doing it for robots. 368 00:22:51,770 --> 00:22:56,090 Ejaaz: And Josh, I guess the final point that I want to make and that we had in this 369 00:22:56,090 --> 00:23:01,870 Ejaaz: essay is Amazon is just really good at understanding what they're good at and 370 00:23:01,870 --> 00:23:05,290 Ejaaz: not treading outside of that line. So what do I mean by that? 371 00:23:05,530 --> 00:23:09,090 Ejaaz: Well, many people think that, okay, if you're a frontier AI company, 372 00:23:09,350 --> 00:23:11,050 Ejaaz: you should have a bleeding edge model. 373 00:23:11,250 --> 00:23:16,210 Ejaaz: The funny part about this is Amazon has arguably the weakest AI model that's 374 00:23:16,210 --> 00:23:19,790 Ejaaz: out there, but it's good for their in-house use. 375 00:23:19,970 --> 00:23:23,010 Ejaaz: So they have a series of models, Josh, and it's called Nova. 376 00:23:23,410 --> 00:23:26,750 Ejaaz: But have you ever used Nova, Josh? Have you ever kind of heard of it? 377 00:23:26,830 --> 00:23:31,050 Josh: No, I have not used. I actually haven't engaged with any Amazon AI yet. This is a new frontier. 378 00:23:31,570 --> 00:23:36,890 Ejaaz: Okay. Well, the reason for that might be because it is a speech-to-speech model, 379 00:23:37,310 --> 00:23:40,430 Ejaaz: meaning that you speak to it and it speaks back to you. 380 00:23:40,510 --> 00:23:43,030 Ejaaz: So you might be like, well, when the hell would I ever do that? 381 00:23:43,330 --> 00:23:47,430 Ejaaz: It's primarily in customer support. That's where they've used their AI model. 382 00:23:47,550 --> 00:23:51,830 Ejaaz: So they trained a very small but hyper-efficient AI model to kind of facilitate 383 00:23:51,830 --> 00:23:55,550 Ejaaz: and backlog all of that kind of stuff. So they've been able to reduce their 384 00:23:55,550 --> 00:23:59,630 Ejaaz: kind of reliance on human customer support for that and save costs in that end. 385 00:23:59,990 --> 00:24:04,270 Ejaaz: The other side of things is Amazon is really good at focusing on enterprise customers. 386 00:24:04,630 --> 00:24:09,730 Ejaaz: And so they produced an AI model software service, which basically allows any 387 00:24:09,730 --> 00:24:15,210 Ejaaz: enterprise to train their own AI model using Amazon's model architecture so 388 00:24:15,210 --> 00:24:18,530 Ejaaz: you can train it on its own proprietary data. Why would you want to do that as an enterprise? 389 00:24:18,650 --> 00:24:21,010 Ejaaz: Well, you have private data, you don't want to give it to OpenAI or you don't 390 00:24:21,010 --> 00:24:23,630 Ejaaz: want to give it to Google, but now you have a private instance where you can 391 00:24:23,630 --> 00:24:26,090 Ejaaz: just run it on Amazon's model product. And that's really cool. 392 00:24:26,430 --> 00:24:30,370 Ejaaz: And that brings me to the final point, Josh, which is like kind of Amazon's 393 00:24:30,370 --> 00:24:34,130 Ejaaz: secret ability here, they've invested in some really big companies. 394 00:24:34,250 --> 00:24:36,110 Ejaaz: In fact, you might've heard of Anthropic. 395 00:24:36,970 --> 00:24:39,530 Ejaaz: Actually, I think you told me this on a previous episode. 396 00:24:39,910 --> 00:24:44,490 Josh: Yeah, a little known fact that most people don't know is that Amazon owns about 397 00:24:44,490 --> 00:24:47,050 Josh: 20%, give or take a few percentage points of Anthropic. 398 00:24:47,470 --> 00:24:49,970 Josh: And it's funny because when I was considering the bear case, 399 00:24:50,090 --> 00:24:54,130 Josh: the reason to be skeptical of Amazon, one of the big things was comparing the 400 00:24:54,130 --> 00:24:57,390 Josh: other major cloud providers. We have Google Cloud Services, which is, 401 00:24:57,510 --> 00:25:00,530 Josh: I mean, a huge entity that runs on Gemini. 402 00:25:00,690 --> 00:25:04,130 Josh: It benefits from Gemini. Then we have Microsoft Azure, which is Microsoft's 403 00:25:04,130 --> 00:25:08,710 Josh: cloud service provider that partners with OpenAI and they receive 100% of the IP from OpenAI. 404 00:25:09,010 --> 00:25:12,550 Josh: But then I was like, well, you know, Amazon actually does have their own big 405 00:25:12,550 --> 00:25:14,990 Josh: dog in their corner, which is Anthropic. 406 00:25:15,210 --> 00:25:18,050 Josh: They have this huge partnership with Anthropic where I'm sure a lot of resources 407 00:25:18,050 --> 00:25:21,690 Josh: are being shared, but Anthropic is also training using these new chips, 408 00:25:22,010 --> 00:25:25,290 Josh: which is fascinating because right now, Now, Anthropic has the most unbelievably 409 00:25:25,290 --> 00:25:27,350 Josh: great coding model in the world. 410 00:25:27,510 --> 00:25:29,830 Josh: So something is working behind the scenes. 411 00:25:30,090 --> 00:25:33,410 Josh: And I guess time will tell us to see how it bleeds out into the rest of the 412 00:25:33,410 --> 00:25:36,350 Josh: industry. But they do have some big guns in their corner. 413 00:25:36,670 --> 00:25:41,030 Ejaaz: Kind of like going on the theme of the bear case, Josh, I have to like put the 414 00:25:41,030 --> 00:25:42,610 Ejaaz: realistic framing on this. 415 00:25:43,700 --> 00:25:47,660 Ejaaz: Amazon's chips are awesome. They're almost as good as NVIDIA. They're 50% cheaper. 416 00:25:47,840 --> 00:25:52,860 Ejaaz: But there's one major constraint, which actually Google faces as well. 417 00:25:53,160 --> 00:25:56,180 Ejaaz: There's a little company in Taiwan called TSMC. 418 00:25:56,680 --> 00:26:01,800 Ejaaz: Some of you might have heard about it. But they require Taiwan Semiconductor 419 00:26:01,800 --> 00:26:06,340 Ejaaz: Manufacturing Company to be able to build the chips for them. 420 00:26:06,480 --> 00:26:08,720 Ejaaz: Google relies on them to build their TPUs as well. 421 00:26:09,120 --> 00:26:14,660 Ejaaz: Can you take a guess at which company owns around 90% of the capacity for TSMC next year. 422 00:26:14,860 --> 00:26:18,060 Josh: Oh, I'm going to guess there's only one correct answer to this question, 423 00:26:18,240 --> 00:26:24,000 Josh: and that is NVIDIA, the owner of all chips and GPUs. 424 00:26:24,380 --> 00:26:29,120 Ejaaz: Yeah, so even if Amazon wanted to, let's say they created a hit product with 425 00:26:29,120 --> 00:26:32,380 Ejaaz: these chips and everyone wanted to use it, they couldn't even fulfill demand 426 00:26:32,380 --> 00:26:34,540 Ejaaz: because NVIDIA holds the stronghold. 427 00:26:34,960 --> 00:26:39,580 Ejaaz: And so they have to wait until the end of 2027 where TSMC would have scaled 428 00:26:39,580 --> 00:26:42,840 Ejaaz: enough capacity at that point to be able to service that. 429 00:26:42,960 --> 00:26:45,600 Ejaaz: So they're going to start to look for alternative providers. 430 00:26:46,140 --> 00:26:49,580 Josh: Yeah, is that a real constraint? Because, okay, so I'm of two minds here where 431 00:26:49,580 --> 00:26:54,680 Josh: Amazon seems to be undervalued just on a relative basis. Their price to earnings ratio is very low. 432 00:26:55,280 --> 00:26:58,840 Josh: Wall Street has basically priced them as break even throughout the course of 433 00:26:58,840 --> 00:27:03,100 Josh: the year, while a lot of the comparable companies have gone up like 80, 90, 100 percent. 434 00:27:03,820 --> 00:27:06,660 Josh: But then I hear things like this and I'm like, well, is this bull case that 435 00:27:06,660 --> 00:27:10,980 Josh: we're laying out actually even possible? Because we do have the TSMC kind of 436 00:27:10,980 --> 00:27:12,260 Josh: monopoly situation with NVIDIA. 437 00:27:12,780 --> 00:27:16,860 Josh: There are alternatives. I know Samsung is working on building another chip architecture. 438 00:27:17,080 --> 00:27:21,420 Josh: Is that really the nail in the coffin? Like, is it possible for them to see 439 00:27:21,420 --> 00:27:26,240 Josh: this upside growth without TSMC? or is it really just relying on TSMC? 440 00:27:26,660 --> 00:27:32,720 Ejaaz: So a consistent trend for NVIDIA that owns the monopoly of capacity on TSMC 441 00:27:32,720 --> 00:27:36,720 Ejaaz: is they want to be able to train frontier AI intelligence. 442 00:27:37,320 --> 00:27:41,840 Ejaaz: Amazon's approach has been very clear. They want to be the cheaper alternative. 443 00:27:41,840 --> 00:27:46,320 Ejaaz: Once you've scaled and built your frontier intelligence, we're the cheaper chip 444 00:27:46,320 --> 00:27:48,620 Ejaaz: for you to use to scale your product in general. 445 00:27:48,940 --> 00:27:52,660 Ejaaz: Inference is where they plan to make most of their money is my suspicion. 446 00:27:53,260 --> 00:27:57,200 Ejaaz: The other way I would address this is, I do think, it's not going to be immediate, 447 00:27:57,340 --> 00:28:01,620 Ejaaz: but eventually, over the next couple of years, companies like Samsung and other 448 00:28:01,620 --> 00:28:05,940 Ejaaz: competitors are eventually going to provide an alternative to TSMC. 449 00:28:06,320 --> 00:28:09,280 Ejaaz: Now, critics will be jumping down my throat on that one because they're going 450 00:28:09,280 --> 00:28:11,360 Ejaaz: to say, well, for the last decade, nothing has changed. 451 00:28:11,680 --> 00:28:15,760 Ejaaz: And I will just respond to you that AI wasn't really a big thing over the last 452 00:28:15,760 --> 00:28:17,940 Ejaaz: decade. It's only over the last couple of years. 453 00:28:18,160 --> 00:28:23,320 Ejaaz: And if big companies like Google, like Amazon are completely constrained because 454 00:28:23,320 --> 00:28:26,520 Ejaaz: of this one company and NVIDIA maintains the market monopoly there, 455 00:28:26,860 --> 00:28:30,320 Ejaaz: it's going to force companies like Samsung to create an alternative. 456 00:28:30,500 --> 00:28:34,360 Ejaaz: We're already seeing this. We had an episode this week or rather last week on 457 00:28:34,360 --> 00:28:40,740 Ejaaz: China copying the key components of a company called ASML to try and fill this gap. 458 00:28:40,920 --> 00:28:45,060 Ejaaz: I just, if I was a betting man, which I am, and I own Amazon stock, 459 00:28:45,300 --> 00:28:50,000 Ejaaz: you know, just putting it out there, I think a competitor is going to come into the ring. 460 00:28:50,470 --> 00:28:54,710 Josh: Yeah, it seems like, well, the interesting thing is Samsung actually beat TSMC 461 00:28:54,710 --> 00:28:58,770 Josh: to the two nanometer chip architecture, which was a really big deal. 462 00:28:58,950 --> 00:29:02,770 Josh: And now they're going to partner with Apple and the largest companies are going 463 00:29:02,770 --> 00:29:03,930 Josh: to get their chips from somewhere else. 464 00:29:04,250 --> 00:29:09,250 Josh: Is it going to be too long? EJS, 2027, we're going to have like AGI on Mars 465 00:29:09,250 --> 00:29:11,170 Josh: by then. So can they survive? 466 00:29:11,630 --> 00:29:15,650 Josh: Like, is that not too long? I don't know. So I guess maybe we could wrap this 467 00:29:15,650 --> 00:29:17,550 Josh: up by kind of where you currently stand. 468 00:29:17,730 --> 00:29:20,370 Josh: How long do you think these things will take to play out? 469 00:29:20,610 --> 00:29:24,010 Josh: How much upside do you think is really possible in the near term? 470 00:29:24,370 --> 00:29:26,290 Josh: How poorly has the market mispriced? 471 00:29:26,410 --> 00:29:29,430 Josh: Is this just like a bubble underwater that is waiting to pop out? 472 00:29:29,670 --> 00:29:33,850 Josh: How does that all look kind of going forward into 2026 based on your understanding of it all? 473 00:29:34,090 --> 00:29:39,210 Ejaaz: Okay, so in short, I think it's bullish because even though they are reliant 474 00:29:39,210 --> 00:29:44,210 Ejaaz: on TSMC's capacity, they are still going to produce in the order of five to 475 00:29:44,210 --> 00:29:48,910 Ejaaz: six million Amazon chips next year. In fact, Google's going to do the same as well. 476 00:29:49,270 --> 00:29:52,010 Ejaaz: And one simple problem remains, Josh. 477 00:29:52,510 --> 00:29:58,450 Ejaaz: There's not enough GPUs, TPUs, or Tranium chips to satisfy demand right now. 478 00:29:58,630 --> 00:30:01,910 Ejaaz: So even with NVIDIA's capacity, they're sold out. 479 00:30:02,170 --> 00:30:05,810 Ejaaz: So companies are looking for other types of chips to fulfill demand. 480 00:30:06,030 --> 00:30:09,830 Ejaaz: That's why OpenAI last week signed a $10 billion deal with Amazon. 481 00:30:09,950 --> 00:30:14,390 Ejaaz: That's why Anthropic signed a 1 million chip deal with Amazon. It's the same thing. 482 00:30:15,050 --> 00:30:18,590 Ejaaz: If Amazon creates all the chips, they're going to end up selling it. 483 00:30:18,810 --> 00:30:22,750 Ejaaz: Now, over the long term, are they going to win? Yes or no? it remains to be seen. 484 00:30:23,250 --> 00:30:26,970 Ejaaz: You know, everyone thinks TSMC has a stronghold. I think there's eventually 485 00:30:26,970 --> 00:30:31,130 Ejaaz: going to be a competitor, which makes Amazon and Google way more competitive to NVIDIA's mode. 486 00:30:31,310 --> 00:30:34,990 Josh: If you build it, they will come. And Amazon is building it. And they're building 487 00:30:34,990 --> 00:30:38,390 Josh: a lot of it. And they have won the least out of everybody. 488 00:30:38,570 --> 00:30:42,750 Josh: So just based on that math alone, there's got to be some hidden upside there. Like they're trading. 489 00:30:43,130 --> 00:30:46,570 Josh: There's such a cash generating business. Everybody who's watching this video 490 00:30:46,570 --> 00:30:48,130 Josh: has interacted with Amazon once in their life. 491 00:30:48,130 --> 00:30:52,110 Josh: It's just it's a great company that will probably stand to benefit from ai when 492 00:30:52,110 --> 00:30:56,130 Josh: and at what scale we don't know but that is part of the fun we will follow it 493 00:30:56,130 --> 00:31:00,030 Josh: along as we go is just are there any parting thoughts before we we wrap up here. 494 00:31:00,030 --> 00:31:04,630 Ejaaz: Actually yes i have one final bit of law we started off the episode critiquing 495 00:31:04,630 --> 00:31:09,370 Ejaaz: amazon stock pricing it's flat one of the main reasons that critics give is 496 00:31:09,370 --> 00:31:13,450 Ejaaz: because they don't really believe in the ceo andy jassy i don't think i agree 497 00:31:13,450 --> 00:31:18,330 Ejaaz: with that he's led the company to kind of like create quarterly earnings upon upon upon a time. 498 00:31:19,010 --> 00:31:23,770 Ejaaz: They're doing really well. They've grown a lot. But there's a small percentage chance, Josh. 499 00:31:24,380 --> 00:31:30,520 Ejaaz: That Jeff Bezos reclaims the throne, similar to how Sergey Brin did so with Google. Why not? 500 00:31:30,700 --> 00:31:36,420 Ejaaz: He's outspokenly said that AI is the most important technological shift over the next decade. 501 00:31:36,600 --> 00:31:40,640 Ejaaz: He started a company that's focused on creating AI lab stuff for robotics and stuff. 502 00:31:40,840 --> 00:31:44,780 Ejaaz: It feels kind of weird for him to just leave his baby behind. 503 00:31:45,040 --> 00:31:47,840 Ejaaz: He knows he already has the vessel there. Why wouldn't he come back? 504 00:31:48,380 --> 00:31:51,760 Josh: Okay, well, it's certainly the same thing as Google. Google had their founders 505 00:31:51,760 --> 00:31:56,240 Josh: disappear. Sergey Brin came back and the stock is up 80% in the 12 months that followed that. 506 00:31:56,500 --> 00:31:59,140 Josh: So there is something, there is a precedent for this happening. 507 00:31:59,580 --> 00:32:03,020 Josh: I mean, we have, Bezos is off, he's building rockets with Blue Origin, he's doing robotics. 508 00:32:03,300 --> 00:32:07,600 Josh: He seems very busy. If he comes back to Amazon, I'm sure that would be bullish to some extent. 509 00:32:07,880 --> 00:32:12,040 Josh: What the extent is, we don't know, but we will be here to cover it as always 510 00:32:12,040 --> 00:32:13,560 Josh: as we go along this journey. 511 00:32:13,760 --> 00:32:17,420 Josh: That will wrap up our episode today on the Amazon Bullcase. If you enjoyed this 512 00:32:17,420 --> 00:32:19,660 Josh: episode, please do not forget to share it with a friend. 513 00:32:19,660 --> 00:32:22,300 Josh: Or if you haven't go rate it five stars on your 514 00:32:22,300 --> 00:32:25,240 Josh: favorite podcast player there is a special ask which 515 00:32:25,240 --> 00:32:28,320 Josh: is actually i'm not even sure this isn't ask this is a value add 516 00:32:28,320 --> 00:32:32,200 Josh: to your life this is the alpha drop that comes two times a week in our newsletter 517 00:32:32,200 --> 00:32:35,920 Josh: every wednesday is a thought piece like the one that we covered today and every 518 00:32:35,920 --> 00:32:39,200 Josh: friday is a short roundup it's five bullets on the most interesting things that 519 00:32:39,200 --> 00:32:43,180 Josh: we found really cool this week in the world of ai and frontier technology so 520 00:32:43,180 --> 00:32:46,660 Josh: if that seems interesting we have the link in our description to go sign up it's a 521 00:32:46,780 --> 00:32:49,560 Josh: stack it's really easy it integrates with all the places you watch you get 522 00:32:49,560 --> 00:32:52,780 Josh: your emails from and i don't know i think it's pretty interesting and 523 00:32:52,780 --> 00:32:55,980 Josh: you do hear the podcast episodes before they come out because that's normally 524 00:32:55,980 --> 00:32:59,540 Josh: where we curate our ideas before presenting them in long form so would highly 525 00:32:59,540 --> 00:33:03,200 Josh: recommend that concludes our first episode this week we still have two more 526 00:33:03,200 --> 00:33:06,280 Josh: to go so stay tuned for that there's a lot of interesting stuff happening and 527 00:33:06,280 --> 00:33:10,140 Josh: thank you so much for watching as always and we will see you guys on the next one see ya.