1 00:00:00,020 --> 00:00:03,300 Josh: There's a chip being built right now in California that the founders of OpenAI 2 00:00:03,300 --> 00:00:07,200 Josh: believe is so important that getting access to it could decide who builds AGI first. 3 00:00:07,420 --> 00:00:11,160 Josh: In fact, the co-founder of OpenAI just testified last week that he and Ilyas 4 00:00:11,160 --> 00:00:15,620 Josh: Zizgever calculated it would take 15 years to build AGI until they discovered 5 00:00:15,620 --> 00:00:17,880 Josh: this hardware that could get them there in just five. 6 00:00:18,040 --> 00:00:22,020 Josh: The name of this company is called Cerebris and they're actually going public tomorrow. 7 00:00:22,300 --> 00:00:27,820 Ejaaz: For the longest time, it has been a unanimous consensus of who the king of AI is. 8 00:00:27,940 --> 00:00:32,520 Ejaaz: It's been NVIDIA and they've own the entire moat on AI, AI chips specifically. 9 00:00:32,760 --> 00:00:35,780 Ejaaz: And this moat has made them famous and pretty valuable. 10 00:00:36,020 --> 00:00:40,040 Ejaaz: They are worth $5.3 trillion, the most valuable company in the world. 11 00:00:40,060 --> 00:00:42,700 Ejaaz: But with success, there's a double-edged sword. 12 00:00:42,740 --> 00:00:46,940 Ejaaz: And it highlighted a bunch of constraints that NVIDIA GPUs couldn't really fix. 13 00:00:47,080 --> 00:00:49,800 Ejaaz: One of these constraints is something known as inference. 14 00:00:49,980 --> 00:00:53,960 Ejaaz: GPUs that NVIDIA create aren't as hyper-optimized towards that. 15 00:00:54,060 --> 00:00:57,340 Ejaaz: And that is the issue that Cerebris focuses on specifically. 16 00:00:57,340 --> 00:01:01,100 Ejaaz: They create these chips which generate lightning-fast inference. 17 00:01:01,220 --> 00:01:04,480 Ejaaz: What this means is if your AI models run on these Cerebrus chips, 18 00:01:04,660 --> 00:01:06,020 Ejaaz: they are lightning quick. 19 00:01:06,180 --> 00:01:09,900 Ejaaz: And in the current AI industry, we're heading towards a world where inference 20 00:01:09,900 --> 00:01:14,620 Ejaaz: is valued and matters way more than AI training for models. 21 00:01:14,740 --> 00:01:18,980 Ejaaz: In fact, it's estimated to be 10 to 50x as large by JP Morgan in their latest report. 22 00:01:19,080 --> 00:01:24,700 Ejaaz: So the point is, if Cerebrus' chips can fulfill that constraint found for inference 23 00:01:24,700 --> 00:01:30,400 Ejaaz: for AI models specifically, this could be the first real threat that NVIDIA faces on their own. 24 00:01:30,600 --> 00:01:32,920 Josh: Okay, so let's talk about briefly, before we get into the IPO, 25 00:01:33,060 --> 00:01:37,060 Josh: what makes Cerebris so special, so different? Because clearly there's some novel architecture here. 26 00:01:37,660 --> 00:01:40,840 Josh: Perhaps we can use this describing it as the pizza problem where 27 00:01:40,840 --> 00:01:43,560 Josh: every chip you've ever owned it kind of starts the same way it takes the 28 00:01:43,560 --> 00:01:46,360 Josh: silicon wafer about the size of a dinner plate here 29 00:01:46,360 --> 00:01:49,240 Josh: that i have for scale so it's this big giant thing 30 00:01:49,240 --> 00:01:52,240 Josh: and it gets cut into a series of tiny little chips that goes into your iphone or 31 00:01:52,240 --> 00:01:55,340 Josh: your ipod or your macbook or whatever it is but cerebris 32 00:01:55,340 --> 00:01:58,300 Josh: decided to question that they said what if you don't cut the pizza what 33 00:01:58,300 --> 00:02:01,040 Josh: if the whole pizza is the chip and that's exactly what they 34 00:02:01,040 --> 00:02:03,920 Josh: did they just took that giant wafer and instead of chopping it 35 00:02:03,920 --> 00:02:06,600 Josh: up into smaller chips they just created it all in one big 36 00:02:06,600 --> 00:02:09,540 Josh: thing and they had this breakthrough that allowed them to fit 37 00:02:09,540 --> 00:02:12,460 Josh: 1.2 trillion transistors on 38 00:02:12,460 --> 00:02:15,300 Josh: a single wafer as opposed to someone like nvidia who currently 39 00:02:15,300 --> 00:02:18,300 Josh: has 21.5 billion and we could see the 40 00:02:18,300 --> 00:02:21,140 Josh: discrepancies on the screen here it is a huge amount of 41 00:02:21,140 --> 00:02:24,440 Josh: additional compute that you could put on a single chip now what does that result 42 00:02:24,440 --> 00:02:27,680 Josh: in because the information doesn't have to travel as far the chip is able to 43 00:02:27,680 --> 00:02:31,360 Josh: access it all so much faster so they put all the memory on the chip all the 44 00:02:31,360 --> 00:02:35,500 Josh: processes on the chip and the result is just significantly faster AI that works 45 00:02:35,500 --> 00:02:38,860 Josh: for the entire stack between inference all the way up to pre training. 46 00:02:39,040 --> 00:02:41,980 Josh: And this is a really huge win for for a lot of the industry. 47 00:02:42,080 --> 00:02:42,700 Josh: It's a big breakthrough. 48 00:02:42,780 --> 00:02:46,420 Ejaaz: And it's not just a theoretical concept, is it either? Like so OpenAI, 49 00:02:46,700 --> 00:02:52,260 Ejaaz: about probably like six months ago, invested $10 billion into their company, 50 00:02:52,260 --> 00:02:55,840 Ejaaz: and they acquired the rights to basically take over all their chip designs and 51 00:02:55,840 --> 00:02:58,420 Ejaaz: use it for OpenAI models specifically. 52 00:02:58,500 --> 00:03:01,600 Ejaaz: And we've seen that manifest in their existing models. 53 00:03:01,740 --> 00:03:05,620 Ejaaz: So OpenAI has this like world leading coding AI model, right? It's called Codex. 54 00:03:05,740 --> 00:03:09,220 Ejaaz: But what most people don't know is there's a fast mode for Codex. 55 00:03:09,280 --> 00:03:11,840 Ejaaz: It's called a GPT Codex Spark. 56 00:03:12,040 --> 00:03:15,320 Ejaaz: And it is absolutely rapid. There's like near zero latency. 57 00:03:15,960 --> 00:03:19,820 Ejaaz: It's running on cerebrous chips. So the concept isn't a concept anymore. 58 00:03:19,820 --> 00:03:22,440 Ejaaz: It's proven at scale to work with these big models. 59 00:03:22,600 --> 00:03:27,400 Ejaaz: And if you are someone that is working on the frontier of software engineering 60 00:03:27,400 --> 00:03:32,620 Ejaaz: and, you know, every iteration cycle, the speed at which you design and ship software matters a lot. 61 00:03:32,680 --> 00:03:35,940 Ejaaz: Or if you're in financial services and you need to get that product out or make 62 00:03:35,940 --> 00:03:38,120 Ejaaz: that trade, you need the fastest chips. 63 00:03:38,300 --> 00:03:42,040 Josh: So these chips are incredibly fast. I mean, we're talking 50 times 64 00:03:42,540 --> 00:03:45,800 Josh: more data on a chip than some of the nvidia chips but the 65 00:03:45,800 --> 00:03:48,660 Josh: result is that it's very expensive these chips are not cheap to 66 00:03:48,660 --> 00:03:51,460 Josh: make the good thing is is that money doesn't seem to be a problem 67 00:03:51,460 --> 00:03:54,200 Josh: when it comes to a lot of these ai companies and open ai 68 00:03:54,200 --> 00:03:57,260 Josh: has committed to what 10 to 20 billion dollars already 69 00:03:57,260 --> 00:04:01,200 Josh: and this is taking us to the headline news of the day which is the ipo the company 70 00:04:01,200 --> 00:04:05,180 Josh: that everyone is super excited about in terms of accelerating the rate that 71 00:04:05,180 --> 00:04:08,820 Josh: we get to agi is actually going public and the numbers are pretty big i mean 72 00:04:08,820 --> 00:04:12,820 Josh: this is a big deal this is the first AI hardware company to really go public 73 00:04:12,820 --> 00:04:15,040 Josh: as we're going on this parabolic trajectory. 74 00:04:15,280 --> 00:04:19,240 Josh: And I think the numbers are going to get pretty staggering in a very quick way. 75 00:04:19,380 --> 00:04:22,860 Ejaaz: Yeah. So to give some context on this, about a week and a half ago, 76 00:04:23,100 --> 00:04:25,960 Ejaaz: Cerebrus announces or basically files for their IPO. 77 00:04:26,120 --> 00:04:29,640 Ejaaz: And there's a prospectus that comes out, right, which explains basically, 78 00:04:29,760 --> 00:04:32,560 Ejaaz: you know, why they're raising and how much specifically they're going to raise. 79 00:04:32,760 --> 00:04:36,600 Ejaaz: A week and a half ago, that number was $3.5 billion that they were going to 80 00:04:36,600 --> 00:04:40,420 Ejaaz: raise. And the The price per share was roughly around $115. 81 00:04:41,380 --> 00:04:47,180 Ejaaz: Something happened immediately after that, though, which is they became 20x oversubscribed. 82 00:04:47,400 --> 00:04:51,240 Ejaaz: So there were 20 times more people that wanted to invest in this thing, 83 00:04:51,360 --> 00:04:56,520 Ejaaz: qualified, institutional, etc., for their IPO, such that they had to revise the terms, Josh. 84 00:04:56,700 --> 00:05:02,380 Ejaaz: So this ended up issuing 2 million more shares at a higher expensive price per share. 85 00:05:02,560 --> 00:05:07,120 Ejaaz: It's 150 bucks. So the total amount that they're raising is actually $4.8 billion, 86 00:05:07,440 --> 00:05:11,720 Ejaaz: call it $5 billion, $1.3 billion above what they were expecting to do, 87 00:05:11,840 --> 00:05:16,100 Ejaaz: which tells me that there is an insatiable amount of demand for this particular IPO. 88 00:05:16,220 --> 00:05:19,300 Ejaaz: Everyone's talking about SpaceX, they're talking about Anthropik and OpenAI's 89 00:05:19,300 --> 00:05:20,840 Ejaaz: potential IPOs later this year. 90 00:05:20,960 --> 00:05:26,740 Ejaaz: But this is one of the real IPOs that show that there's a huge demand in the West for AI companies. 91 00:05:26,940 --> 00:05:29,860 Ejaaz: And it's exciting to see because this is the first of many. 92 00:05:29,860 --> 00:05:32,880 Josh: Yeah, well, we have Cerebris kicking things off. Then we have the rumor for 93 00:05:32,880 --> 00:05:37,020 Josh: SpaceX, OpenAI, Anthropik, Databricks, perhaps even Stripe. 94 00:05:37,140 --> 00:05:42,940 Josh: That's a combined potential $3.6 trillion in market cap that could be entering the market. 95 00:05:43,060 --> 00:05:46,120 Josh: And we have this hot ball of money. We just filmed an amazing episode last week 96 00:05:46,120 --> 00:05:49,920 Josh: about the AI stack and where every company kind of lines up and where the opportunities 97 00:05:49,920 --> 00:05:52,540 Josh: are for this new era of AI investing. 98 00:05:52,860 --> 00:05:57,100 Josh: And this sits very squarely right in the hardware era, right next to someone 99 00:05:57,100 --> 00:05:59,420 Josh: kind of like NVIDIA, right? Like they're not quite Nvidia's size. 100 00:05:59,560 --> 00:06:02,980 Josh: They're nowhere near Nvidia's size, but they're competing on the same kind of 101 00:06:02,980 --> 00:06:06,380 Josh: level, the same type of architecture that a company like Nvidia is just at a 102 00:06:06,380 --> 00:06:09,060 Josh: very different angle using these huge pizza-sized wafers. 103 00:06:09,220 --> 00:06:12,900 Ejaaz: Yeah, so what I put and want to make on the Nvidia thing specifically is, 104 00:06:13,180 --> 00:06:17,560 Ejaaz: okay, so there's one framing of this where if the Cerebrus IPO is successful, 105 00:06:17,800 --> 00:06:23,760 Ejaaz: it's the first real signal that Nvidia's monopoly on AI chips is actually going 106 00:06:23,760 --> 00:06:24,840 Ejaaz: to be threatened, right? 107 00:06:24,960 --> 00:06:27,900 Ejaaz: And this might sound like a crazy thing to say, right? It's a small company, 108 00:06:27,900 --> 00:06:31,280 Ejaaz: they're raising $5 billion versus NVIDIA's $5.3 trillion. 109 00:06:32,230 --> 00:06:36,590 Ejaaz: There's another company which proved that NVIDIA's mode might also be shakable, right? 110 00:06:36,770 --> 00:06:40,010 Ejaaz: And that is Google last week, who became the most valuable company. 111 00:06:40,090 --> 00:06:42,870 Ejaaz: I don't think they're currently the most valuable company, but they beat NVIDIA. 112 00:06:43,010 --> 00:06:46,830 Ejaaz: And the reason why they did that was because they released these brand new TPUs, 113 00:06:46,890 --> 00:06:49,690 Ejaaz: their version of their own GPUs, which they train their Gemini model on, 114 00:06:49,930 --> 00:06:53,910 Ejaaz: which so many customers, including Anthropic and OpenAI, want to use to train 115 00:06:53,910 --> 00:06:54,890 Ejaaz: and inference their own model. 116 00:06:55,010 --> 00:07:00,690 Ejaaz: So the point is, the ground is shifting beneath NVIDIA, even though they have 117 00:07:00,690 --> 00:07:03,390 Ejaaz: like decades and decades worth of experience building these GPUs, 118 00:07:03,450 --> 00:07:08,870 Ejaaz: there might still be valid, compatible threats against their particular mode. 119 00:07:09,230 --> 00:07:12,090 Ejaaz: The other thing I want to say is, and I mentioned this earlier, 120 00:07:12,490 --> 00:07:14,130 Ejaaz: inference versus training. 121 00:07:14,350 --> 00:07:19,190 Ejaaz: For the longest time, people have said AI training is where all the money is going to be. 122 00:07:19,290 --> 00:07:22,610 Ejaaz: And that's where NVIDIA has lived and breathed over the last decade. 123 00:07:22,790 --> 00:07:27,650 Ejaaz: But what has shifted recently is what matters more is what happens after the training. 124 00:07:27,830 --> 00:07:31,010 Ejaaz: How many people are going to be using agents that query the models. 125 00:07:31,410 --> 00:07:34,250 Ejaaz: Inference is where the money is going to be made, where the value of the opportunity 126 00:07:34,250 --> 00:07:37,510 Ejaaz: is going to be huge. And Cerebrus plays directly in that point. 127 00:07:37,930 --> 00:07:41,530 Ejaaz: NVIDIA actually proved this because they acquired another company called Grok 128 00:07:41,530 --> 00:07:44,950 Ejaaz: for $20 billion, which does something similar to what Cerebrus did. 129 00:07:45,070 --> 00:07:49,070 Ejaaz: So it's validating the fact that Cerebrus is going to be valuable and now the 130 00:07:49,070 --> 00:07:50,610 Ejaaz: public can get exposure to it. It's cool. 131 00:07:50,710 --> 00:07:54,490 Josh: And I have some older benchmarks about how fast Cerebrus runs versus the rest 132 00:07:54,490 --> 00:07:59,070 Josh: of the world. uh taking a model like llama for maverick uh cerebris was delivering 133 00:07:59,070 --> 00:08:04,610 Josh: at 2500 tokens per second which is more than two times nvidia's flagship blackwell chip 134 00:08:04,980 --> 00:08:09,740 Josh: And then if you lower it down to Llama 3.1, which is a $70 billion parameter 135 00:08:09,740 --> 00:08:12,840 Josh: model, it was 20 times faster than NVIDIA. 136 00:08:13,000 --> 00:08:16,060 Josh: So the actual rate that it's able to deliver these tokens is huge. 137 00:08:16,180 --> 00:08:19,220 Josh: And one of the most important things that we're seeing right now is kind of 138 00:08:19,220 --> 00:08:22,160 Josh: reasoning, chain of thought, deep thinking of these agents. 139 00:08:22,300 --> 00:08:24,940 Josh: And when you're able to deliver tokens at a much faster rate, 140 00:08:25,100 --> 00:08:28,480 Josh: you can get from the question to the answer much quicker. 141 00:08:28,680 --> 00:08:31,360 Josh: One of the new things that we've seen in Codex recently that has become super 142 00:08:31,360 --> 00:08:33,580 Josh: popular is the backslash goal command. 143 00:08:33,580 --> 00:08:36,700 Josh: Where you can set it a command it'll go off and think for 24 144 00:08:36,700 --> 00:08:39,740 Josh: 36 72 hours of time to 145 00:08:39,740 --> 00:08:43,320 Josh: collect enough information to get to a single answer this can 146 00:08:43,320 --> 00:08:46,280 Josh: make that process significantly quicker and therefore make agents 147 00:08:46,280 --> 00:08:50,980 Josh: much more capable so the reasoning speed in terms of inference in terms of delivering 148 00:08:50,980 --> 00:08:55,640 Josh: reasoning it's so important and although currently i believe fast mode on chat 149 00:08:55,640 --> 00:08:59,540 Josh: gpt is about 1.5 times the cost you have to suspect that that cost is going 150 00:08:59,540 --> 00:09:03,480 Josh: to go down at some point and that makes Cerebris a really compelling company. 151 00:09:03,700 --> 00:09:06,100 Josh: And I think that's what OpenA I saw when they made this huge investment in them. 152 00:09:06,260 --> 00:09:09,140 Josh: Now, Ijaz, I have to ask you, what are you thinking about this IPO? 153 00:09:09,260 --> 00:09:11,140 Josh: Is this something that you're interested in investing in? 154 00:09:11,800 --> 00:09:15,280 Ejaaz: It's a tricky one because my default answer is yes. 155 00:09:15,500 --> 00:09:19,980 Ejaaz: But if I want to like touch grass for a second to get off my eye holes... 156 00:09:20,920 --> 00:09:26,240 Ejaaz: The IPO is kind of already happening. It's just happening in the private market, 157 00:09:26,440 --> 00:09:28,280 Ejaaz: if that makes sense. Like, think about this, right? 158 00:09:28,560 --> 00:09:33,280 Ejaaz: NVIDIA, when they IPO'd, I believe was, they IPO'd at around $800 million. 159 00:09:33,840 --> 00:09:36,560 Ejaaz: That's million. It's under a billion dollars, right? 160 00:09:36,840 --> 00:09:40,160 Ejaaz: And currently we have like Cerebras, which, you know, they have one major customer 161 00:09:40,160 --> 00:09:44,500 Ejaaz: as OpenAI. They've proved the concept fair well, but they're raising $5 billion. 162 00:09:44,760 --> 00:09:49,080 Ejaaz: And this is all pre-IPO. So there's a lot of people that are making a lot of 163 00:09:49,080 --> 00:09:51,680 Ejaaz: money. we see this with the Anthropic secondary shares, OpenAI, 164 00:09:51,820 --> 00:09:55,500 Ejaaz: the same thing with SpaceX as well, before the actual IPO happened. 165 00:09:55,600 --> 00:09:59,300 Ejaaz: So the skeptics are saying, this is just going to be a retail liquidity event 166 00:09:59,300 --> 00:10:03,040 Ejaaz: where they're going to dump on retail and no one's going to really make any money after that. 167 00:10:03,200 --> 00:10:07,080 Ejaaz: I don't think there's a zero chance of that happening, but there is a chance. 168 00:10:07,300 --> 00:10:12,500 Ejaaz: Now, the bull case for this IPO is one specific reason, which is there is going 169 00:10:12,500 --> 00:10:15,780 Ejaaz: to be an insatiable demand for two things, Josh. 170 00:10:16,120 --> 00:10:21,500 Ejaaz: One is I need to use this AI to do things for me, whether it's agents or whether it's me prompting. 171 00:10:21,580 --> 00:10:24,960 Ejaaz: Actually, I'm more bullish on the agent side, but we can tap into that as a second thing. 172 00:10:25,380 --> 00:10:32,240 Ejaaz: Second thing is Cerebrus's chips has something very unique that NVIDIA chips can't get enough of. 173 00:10:32,340 --> 00:10:37,400 Ejaaz: And that's something called memory, specifically static random access memory. 174 00:10:37,540 --> 00:10:41,900 Ejaaz: It's a very unique type of memory, which Cerebrus installs into their chips. 175 00:10:42,020 --> 00:10:44,260 Ejaaz: And it's actually what makes them incredibly unique. 176 00:10:44,980 --> 00:10:49,820 Ejaaz: If you look at all the typical 99.9% of GPUs that people use, 177 00:10:49,980 --> 00:10:52,020 Ejaaz: they use this thing called high bandwidth memory. 178 00:10:52,160 --> 00:10:54,760 Ejaaz: It's composed of something known as dynamic random access memory. 179 00:10:54,860 --> 00:10:58,500 Ejaaz: It's kind of like the sister to static random access memory. It's cheaper. 180 00:10:58,720 --> 00:11:02,180 Ejaaz: It's more scalable. It's the sexier thing right now. It's what all the major 181 00:11:02,180 --> 00:11:05,060 Ejaaz: memory manufacturers have focused on for the last decade. 182 00:11:05,220 --> 00:11:08,760 Ejaaz: And that's what's being installed by default into these GPUs. 183 00:11:08,840 --> 00:11:13,860 Ejaaz: But there's one issue. They don't deliver the information that the memory stores 184 00:11:13,860 --> 00:11:17,860 Ejaaz: very effectively or quickly enough to allow these models to run very quickly. 185 00:11:18,080 --> 00:11:23,120 Ejaaz: Now, one solution to this is static random access memory, which Cerebris has 186 00:11:23,120 --> 00:11:25,580 Ejaaz: focused on for their entirety of their existence. 187 00:11:25,600 --> 00:11:29,740 Ejaaz: And they have perfected the design and installation method for their particular 188 00:11:29,740 --> 00:11:32,200 Ejaaz: chips. There's a reason why it looks so much larger. 189 00:11:32,320 --> 00:11:36,120 Ejaaz: You know, you used that dinner plate just now, Josh, versus the NVIDIA chips. 190 00:11:36,240 --> 00:11:39,920 Ejaaz: It is more expensive. You mentioned that the cost to use the things is more expensive. 191 00:11:40,020 --> 00:11:45,400 Ejaaz: But think about how much more value you could get by using quicker turns of your AI model. 192 00:11:45,540 --> 00:11:49,840 Ejaaz: You could end up having a super fast quant algorithm to trade your fund or to 193 00:11:49,840 --> 00:11:53,580 Ejaaz: deliver access to a product for millions of your different customers. 194 00:11:53,820 --> 00:11:59,660 Ejaaz: And that in itself might be worth billions and billions more than having a slower memory component. 195 00:11:59,800 --> 00:12:02,940 Ejaaz: So that is the clear distinction between them and what makes me the most bullish 196 00:12:02,940 --> 00:12:05,100 Ejaaz: on Cerebra's IPO specifically. 197 00:12:05,240 --> 00:12:08,520 Josh: Yeah, as I was reading about it earlier this morning, the SRAM versus DRAM conversation, 198 00:12:08,520 --> 00:12:12,020 Josh: one of the easiest ways that i was able to kind of differentiate between the 199 00:12:12,020 --> 00:12:16,820 Josh: two is an sram is much more similar to an ssd a solid state drive whereas dram 200 00:12:16,820 --> 00:12:20,860 Josh: is kind of similar to a htd a hard drive a hard disk drive something that has 201 00:12:20,860 --> 00:12:24,900 Josh: an actual needle where it has a lot more capacity but it's a bit slower meanwhile sram 202 00:12:25,400 --> 00:12:30,540 Josh: So long as there is power provided to it, it will retain that memory and information indefinitely. 203 00:12:30,780 --> 00:12:34,600 Josh: So in DRAM, one of the big problems that happens is there is this cache memory 204 00:12:34,600 --> 00:12:38,600 Josh: and it has to continue to refresh itself over and over because it's a very short shelf life. 205 00:12:38,940 --> 00:12:43,420 Josh: With SRAM, that shelf life is infinite and it can be stored as long as there's 206 00:12:43,420 --> 00:12:46,660 Josh: power and it can be recalled very quickly without that constant refresh. 207 00:12:46,840 --> 00:12:50,140 Josh: And that's the huge breakthrough. So it's a really compelling product that I 208 00:12:50,140 --> 00:12:55,560 Josh: think is very impressive and very necessary in this world. The only real other 209 00:12:55,560 --> 00:12:58,620 Josh: GPU architecture we've seen that has had success is Grok. 210 00:12:59,120 --> 00:13:02,860 Josh: And we saw what happened with them. They had an amazing exit and they are working 211 00:13:02,860 --> 00:13:04,180 Josh: now directly with NVIDIA. 212 00:13:04,300 --> 00:13:09,000 Josh: So this is the direct comparison. And for that reason, I think I'm pretty bullish 213 00:13:09,000 --> 00:13:09,960 Josh: on the company because... 214 00:13:10,940 --> 00:13:15,800 Josh: Essentially, Cerebrus is tied to OpenAI success. We know that both founders 215 00:13:15,800 --> 00:13:16,920 Josh: are actually invested personally. 216 00:13:17,960 --> 00:13:21,420 Josh: Greg Brockman and Sam Altman are both personally invested in the company. 217 00:13:21,860 --> 00:13:25,640 Josh: OpenAI is invested in the company. And OpenAI has every incentive in the world 218 00:13:25,640 --> 00:13:28,280 Josh: to scale with them and to grow with Cerebrus. 219 00:13:28,840 --> 00:13:33,120 Josh: That aligned with the fact that this is the first AI IPO we've really had in 220 00:13:33,120 --> 00:13:34,460 Josh: this chain that's going to kick things off. 221 00:13:34,580 --> 00:13:37,660 Josh: I expect there to be a lot of excitement, a lot of enthusiasm so long 222 00:13:37,660 --> 00:13:41,000 Josh: as the market continues to go up so long as these this token demand continues 223 00:13:41,000 --> 00:13:45,660 Josh: i see no reason why cerebrus wouldn't follow in that path and actually be somewhat 224 00:13:45,660 --> 00:13:49,120 Josh: successful i think i'd love for everyone who's watching to share their takes 225 00:13:49,120 --> 00:13:53,960 Josh: are you investing in the ipo why why not this is uh it's going to be a big conversation 226 00:13:53,960 --> 00:13:55,780 Josh: over the next coming weeks as we go through this another 227 00:13:55,780 --> 00:13:59,560 Ejaaz: Thing is like um okay it's one thing having a good product right it's another 228 00:13:59,560 --> 00:14:03,540 Ejaaz: thing like getting the right distribution now them signing this like 20 billion 229 00:14:03,540 --> 00:14:08,440 Ejaaz: uh partnership chip with OpenAI is good, right? You get access to, what do they have now? 230 00:14:08,560 --> 00:14:12,540 Ejaaz: Almost like a billion weekly active users or something crazy like that. 231 00:14:13,260 --> 00:14:19,420 Ejaaz: The other thing is their chips are integrated into Amazon's Bedrock platform. 232 00:14:19,640 --> 00:14:23,800 Ejaaz: I don't know if you knew this, Josh, but Amazon basically like owns the entire 233 00:14:23,800 --> 00:14:29,260 Ejaaz: enterprise compute market because they give access to every enterprise that 234 00:14:29,260 --> 00:14:34,340 Ejaaz: wants to spin up an AWS instance or get access to specific tools through AWS, 235 00:14:34,620 --> 00:14:35,860 Ejaaz: their Amazon web service. 236 00:14:36,240 --> 00:14:40,100 Ejaaz: Now, they run Cerebrus chips themselves there. So Cerebrus now has a distribution 237 00:14:40,100 --> 00:14:41,280 Ejaaz: mode through that, right? 238 00:14:41,360 --> 00:14:45,480 Ejaaz: So the point is, distribution is solved, the ingenuity of going zero to one 239 00:14:45,480 --> 00:14:49,940 Ejaaz: with SRAM is also solved, and they have the consumer mode being funneled through OpenAI. 240 00:14:50,060 --> 00:14:54,360 Ejaaz: So all of this sounds incredibly bullish, but I do want to touch on the bear 241 00:14:54,360 --> 00:14:59,220 Ejaaz: case for a very brief moment, right and i mentioned the first one um earlier 242 00:14:59,220 --> 00:15:00,240 Ejaaz: on which is the valuation. 243 00:15:00,770 --> 00:15:03,910 Ejaaz: Is pretty steep, right? The fact that they were 20x oversubscribed and they 244 00:15:03,910 --> 00:15:08,030 Ejaaz: just revised the terms in the last five days to give even more access is kind 245 00:15:08,030 --> 00:15:10,950 Ejaaz: of, I'm just like, so typically, just for those of you who don't know, 246 00:15:11,330 --> 00:15:19,950 Ejaaz: for the average IPO over the last, I think, five years, the price on IPO pops up between 30 to 80%. 247 00:15:19,950 --> 00:15:23,950 Ejaaz: So if you're a retailer, right, that wants to get involved here and like buy 248 00:15:23,950 --> 00:15:26,970 Ejaaz: this thing, you might be buying it for like a significant premium. 249 00:15:27,110 --> 00:15:30,030 Ejaaz: And these things tend to dump a little bit after the first couple of days. 250 00:15:30,210 --> 00:15:34,230 Ejaaz: And then, you know, whatever, it figures itself out. So the valuation is pretty steep. 251 00:15:34,490 --> 00:15:39,530 Ejaaz: The other thing I want to say is that OpenAI is also building their own silicon. 252 00:15:39,830 --> 00:15:44,450 Ejaaz: They have a deal with Broadcom, MediaTek, and a bunch of these other intermediaries 253 00:15:44,450 --> 00:15:46,210 Ejaaz: to design their own chip. 254 00:15:46,370 --> 00:15:51,310 Ejaaz: And Sam has specifically said that one of the major cases that he's trying to solve is inference. 255 00:15:51,590 --> 00:15:54,570 Ejaaz: So I'm kind of looking at this and I'm like, okay, so you've invested in this 256 00:15:54,570 --> 00:15:57,050 Ejaaz: company, you're going, you know, you're running their chips, 257 00:15:57,210 --> 00:16:00,530 Ejaaz: it's doing the thing, but you're also building your own custom silicon. 258 00:16:00,710 --> 00:16:03,310 Ejaaz: Now, granted, that's probably going to take years to try and figure out. 259 00:16:03,430 --> 00:16:06,990 Ejaaz: But is the idea that you eventually acquire kind of Cerebrus, 260 00:16:07,090 --> 00:16:10,090 Ejaaz: but they're IPOing, so I guess you can't really do this and merge them together? 261 00:16:10,330 --> 00:16:13,750 Ejaaz: Or are you going to eventually ditch them and they're like an intermediary solution? 262 00:16:13,870 --> 00:16:17,530 Ejaaz: And so Cerebrus might end up like losing this customer going down the road. 263 00:16:17,630 --> 00:16:19,670 Ejaaz: I don't know, but those are the two things that sound out for me. 264 00:16:19,970 --> 00:16:22,990 Josh: That's fair. Yeah, there's also they're trying to compete against the CUDA moat, 265 00:16:23,130 --> 00:16:26,290 Josh: which is i mean very powerful and very very strong 266 00:16:26,290 --> 00:16:29,330 Josh: as we know they also are i mean they're trading devaluation point 267 00:16:29,330 --> 00:16:33,410 Josh: 51 times revenue a lot of these larger ai companies are trading actually at 268 00:16:33,410 --> 00:16:37,650 Josh: fairly low multiples right now a lot of the major players in the mag 7 are are 269 00:16:37,650 --> 00:16:41,990 Josh: no more than like a 20 times revenue this is at 51 and there's also one interesting 270 00:16:41,990 --> 00:16:45,510 Josh: component where there is actually a memory ceiling currently on the chips where 271 00:16:45,510 --> 00:16:48,070 Josh: they're able to fit 44 gigabytes on a chip and currently, 272 00:16:48,230 --> 00:16:51,550 Josh: that's unbelievable, but it requires a lot more 273 00:16:52,200 --> 00:16:55,380 Josh: Innovation if they're going to continue to scale to provide 274 00:16:55,380 --> 00:16:58,460 Josh: inference for multi-modality and for video 275 00:16:58,460 --> 00:17:01,340 Josh: models or for ultra long context there's going to need to be 276 00:17:01,340 --> 00:17:07,920 Josh: continued innovation and i think that the ceiling is not quite as clear like 277 00:17:07,920 --> 00:17:11,200 Josh: we're not quite clear where this can go to we know it's incredible now i'm not 278 00:17:11,200 --> 00:17:13,880 Josh: sure people really know where that's going to go to that's perhaps a longer 279 00:17:13,880 --> 00:17:17,020 Josh: term problem that's not something that would really impact the company short 280 00:17:17,020 --> 00:17:19,200 Josh: term but it is something to be noteworthy of. 281 00:17:19,320 --> 00:17:22,300 Josh: I mean, this is, it's an amazing technology. It's like in your computer, 282 00:17:22,300 --> 00:17:25,200 Josh: if you have your RAM and your hard drive, 283 00:17:25,730 --> 00:17:28,910 Josh: as separate components, this merges them into one. And now everything is so 284 00:17:28,910 --> 00:17:29,770 Josh: much faster because of it. 285 00:17:29,830 --> 00:17:33,210 Josh: And the unlock that has for generating tokens, which is the single most important 286 00:17:33,210 --> 00:17:34,690 Josh: thing that we do, is huge. 287 00:17:34,830 --> 00:17:37,750 Josh: So I think as a company, Cerebris is something to be very excited about. 288 00:17:38,050 --> 00:17:42,010 Josh: This is going to be a fascinating use case or a fascinating case study as they 289 00:17:42,010 --> 00:17:44,590 Josh: go public this week to see what that type of impact has on the market. 290 00:17:44,930 --> 00:17:47,790 Josh: And overall, I'm feeling really excited for some bullish innovation. 291 00:17:47,970 --> 00:17:50,790 Josh: I mean, Cerebris is a huge unlock in the world of AI. 292 00:17:51,230 --> 00:17:53,990 Josh: Sam and Greg saw it first. The rest of the world is going to see it this week 293 00:17:53,990 --> 00:17:56,710 Josh: as they go public. And we'll see what the downstream effects of it are. 294 00:17:56,830 --> 00:18:01,350 Ejaaz: I mean, it's the first domino of many, right? I think another obvious reason 295 00:18:01,350 --> 00:18:04,790 Ejaaz: why it's getting a lot of attention is like, it's AI summer, right? 296 00:18:04,850 --> 00:18:07,210 Ejaaz: We've got a bunch of cool IPOs that are coming up. 297 00:18:07,690 --> 00:18:10,970 Ejaaz: You know, I've got a graphic here, which is actually severely outdated. 298 00:18:11,150 --> 00:18:15,730 Ejaaz: It's funny, it values Anthropik's IPO as being around $230 billion. 299 00:18:15,990 --> 00:18:18,930 Ejaaz: And I'm like, have you looked at the secondary market for their shares recently? 300 00:18:19,070 --> 00:18:22,010 Ejaaz: They're selling at like at a trillion dollar valuation. People are literally 301 00:18:22,010 --> 00:18:26,730 Ejaaz: leasing their homes out to be able to like fund like a purchase of Anthropic secondary shares. 302 00:18:26,890 --> 00:18:30,050 Ejaaz: It's pretty crazy out there. But the point being is you've got SpaceX. 303 00:18:30,550 --> 00:18:33,630 Ejaaz: You've got Open Air, you've got ByteDance from China, you've got Anthropic that 304 00:18:33,630 --> 00:18:35,850 Ejaaz: are all planning to IPO this year. 305 00:18:36,150 --> 00:18:39,350 Ejaaz: There is a lot of money that's going to be entering the market. 306 00:18:39,550 --> 00:18:43,090 Ejaaz: Some of it will be net new coming in from retail, coming in from funds that 307 00:18:43,090 --> 00:18:44,370 Ejaaz: don't typically invest in these things. 308 00:18:44,550 --> 00:18:48,130 Ejaaz: But a lot of it is going to come from alternative types of companies. 309 00:18:48,290 --> 00:18:51,530 Ejaaz: Now, we've seen signals of where this is going to come from. 310 00:18:51,630 --> 00:18:55,630 Ejaaz: Whenever, for example, whenever Anthropic announces a new product or feature 311 00:18:55,630 --> 00:19:00,550 Ejaaz: that replaces a certain segment or industrial sector, you see the top market 312 00:19:00,550 --> 00:19:04,870 Ejaaz: cap stocks of those sectors dump as they've tweeted the announcement out, right? 313 00:19:04,990 --> 00:19:08,390 Ejaaz: So it might be like SaaS companies are cooked or whatever that might be, 314 00:19:08,490 --> 00:19:11,310 Ejaaz: but it'll be interesting to see the market dynamics because I think there'll 315 00:19:11,310 --> 00:19:14,150 Ejaaz: be a lot of money sloshing around into these big IPOs. 316 00:19:14,250 --> 00:19:16,770 Ejaaz: And then the major question that skeptics have is, 317 00:19:17,060 --> 00:19:19,920 Ejaaz: They don't think there's enough money. And this might be, you know, 318 00:19:20,060 --> 00:19:23,360 Ejaaz: the popping of the bubble metaphorically or financially. 319 00:19:23,620 --> 00:19:26,720 Ejaaz: We might see a bunch of these like crash after the liquidity event. 320 00:19:26,940 --> 00:19:30,220 Ejaaz: That's more of the bad case, but I'm choosing to remain bullish on this because 321 00:19:30,220 --> 00:19:33,400 Ejaaz: I think the demand is pretty insatiable, right? 322 00:19:33,560 --> 00:19:38,680 Ejaaz: Today, OpenAI announced like a huge partnership with enterprise for deployed engineers. 323 00:19:39,020 --> 00:19:42,680 Ejaaz: Anthropic did the same thing five days ago. So they're generating highest revenues. 324 00:19:42,880 --> 00:19:47,800 Ejaaz: Like Anthropic hit 45 billion ARR. they're aiming to hit 100 by the end of the year. 325 00:19:47,940 --> 00:19:51,020 Ejaaz: These aren't made up numbers. So, you know, I'm excited about it. 326 00:19:51,500 --> 00:19:54,900 Josh: Unbelievable businesses. And yeah, I think to your point, there may be some 327 00:19:54,900 --> 00:19:57,760 Josh: liquidity issues as we get to IPO number four, five, and six. 328 00:19:58,020 --> 00:20:01,860 Josh: This is the smallest of all or among the smallest of all of them that's happening. 329 00:20:01,880 --> 00:20:05,780 Josh: So I'm not exactly concerned about that yet, but it's going to be a crazy year. 330 00:20:05,940 --> 00:20:10,160 Josh: And this will certainly set some sort of a precedent for how the market considers IPOs. 331 00:20:10,320 --> 00:20:12,340 Ejaaz: Are you buying some of this tomorrow? I have to ask you. 332 00:20:12,340 --> 00:20:16,640 Josh: You know what? I think it'll depend on how it trades. I probably won't buy it 333 00:20:16,640 --> 00:20:17,900 Josh: at the open. Maybe wait a little bit. 334 00:20:18,180 --> 00:20:20,360 Ejaaz: You're going to wait for that 30 to 80% premium? 335 00:20:22,490 --> 00:20:25,810 Josh: I think the answer is yes. Like, I feel like I should, I should participate 336 00:20:25,810 --> 00:20:29,870 Josh: at least in a small amount in an IPO like this, because even if it is overpriced 337 00:20:29,870 --> 00:20:32,130 Josh: on the short term, it'll reprice itself over the next few weeks. 338 00:20:32,190 --> 00:20:37,470 Josh: And I feel like six months from now, this company will produce better products than it does today. 339 00:20:37,690 --> 00:20:42,090 Josh: And the same will also be true 12 months from now. So I'm not much of a trader. I'm more of an investor. 340 00:20:42,250 --> 00:20:45,250 Josh: This is a compelling investment. And I could also see a world in which it does 341 00:20:45,250 --> 00:20:49,190 Josh: get acquired or a majority of it does wind up getting bought out by another 342 00:20:49,190 --> 00:20:50,910 Josh: company. It's just incredibly important infrastructure. 343 00:20:51,150 --> 00:20:54,790 Josh: It seems valuable. We saw what happened with Grok. We might have an opportunity 344 00:20:54,790 --> 00:20:56,570 Josh: to actually participate in it again with Cerebris. 345 00:20:56,770 --> 00:20:59,490 Josh: I'm feeling bullish. I'm feeling like I want to get in the mix. 346 00:20:59,550 --> 00:21:00,950 Josh: I think I want to get involved a little bit. 347 00:21:01,590 --> 00:21:07,810 Ejaaz: Same. Hey, I'm down to buy NVIDIA at under a $10 billion valuation. 348 00:21:08,630 --> 00:21:13,650 Ejaaz: If it even achieves one of 10% of NVIDIA's market cap, that would be a win. 349 00:21:13,890 --> 00:21:15,430 Ejaaz: So yeah, I'm excited to see how this plays out. 350 00:21:15,750 --> 00:21:19,030 Josh: Yeah. Assuming that thesis holds up where we continue used to need tokens and 351 00:21:19,030 --> 00:21:22,650 Josh: those tokens were able to generate a profit, the people who can distribute and 352 00:21:22,650 --> 00:21:24,650 Josh: produce those tokens faster than everyone else will win. 353 00:21:24,770 --> 00:21:28,110 Josh: And I think that is a valid enough thesis to have me convinced. 354 00:21:28,310 --> 00:21:32,350 Josh: But to the listener, to everyone out there who has just watched 20 minutes of 355 00:21:32,350 --> 00:21:33,370 Josh: this episode, do you agree? 356 00:21:33,690 --> 00:21:36,490 Josh: Are you participating in Cerebris? Is this overpriced? Is it underpriced? 357 00:21:36,570 --> 00:21:39,030 Josh: Is it stupid? Should we even be investing or considering? 358 00:21:39,410 --> 00:21:41,110 Josh: Let us know in the comments down below. If you enjoyed this video, 359 00:21:41,290 --> 00:21:44,470 Josh: please don't forget to share with your friends. Ija, any final parting thoughts? 360 00:21:44,690 --> 00:21:48,570 Ejaaz: I'm actually curious whether the listeners of the show enjoyed this episode, 361 00:21:48,950 --> 00:21:51,810 Ejaaz: we do various different types of investment ones, the ones that we did last 362 00:21:51,810 --> 00:21:55,490 Ejaaz: week on the AI info stack got so much attention. 363 00:21:55,650 --> 00:21:58,470 Ejaaz: And we're going to do a follow up episode on that just we heard we read all 364 00:21:58,470 --> 00:21:59,450 Ejaaz: your comments, we're going to do it. 365 00:21:59,730 --> 00:22:02,050 Ejaaz: But on something like this, like, is this something that's interesting? 366 00:22:02,170 --> 00:22:04,990 Ejaaz: Do you want to hear about upcoming IPOs? I'm curious to hear from you guys. 367 00:22:05,110 --> 00:22:06,850 Ejaaz: But aside for that, I think we're all good.