1 00:00:00,020 --> 00:00:04,160 Josh: Everyone's asking, are we in a bubble? When is the bubble going to pop? 2 00:00:04,320 --> 00:00:08,120 Josh: How big can AI get before it all comes crumbling down? 3 00:00:08,340 --> 00:00:15,260 Josh: And yesterday, the OpenAI CFO may have delivered some hints as to how and why this happens. 4 00:00:15,660 --> 00:00:19,260 Josh: She used the word, and I'm ashamed to say it out loud, this feels like a bad 5 00:00:19,260 --> 00:00:20,600 Josh: word, she used the word backstop. 6 00:00:20,940 --> 00:00:24,540 Josh: And if you're familiar with the word backstop, it's what a lot of banks were 7 00:00:24,540 --> 00:00:29,580 Josh: experiencing in 2008, where the government kind of sits there to fund any sort 8 00:00:29,580 --> 00:00:32,780 Josh: of negative repercussions that happen as a result of this build out. 9 00:00:32,900 --> 00:00:36,040 Josh: And then after seeing this interview, it kind of begs the question, are these circular 10 00:00:36,680 --> 00:00:39,680 Josh: economics we've been seeing among NVIDIA, Microsoft, AMD, 11 00:00:39,980 --> 00:00:42,900 Josh: is it really a capitalist efficient thing or 12 00:00:42,900 --> 00:00:45,800 Josh: is this is are we really starting to see early signs of a 13 00:00:45,800 --> 00:00:48,700 Josh: bubble starting to burst so we're going to get into all that we are also talking 14 00:00:48,700 --> 00:00:51,560 Josh: about a deal that was just confirmed with apple where 15 00:00:51,560 --> 00:00:55,180 Josh: they are paying google a tremendous amount of money to actually offload their 16 00:00:55,180 --> 00:00:58,200 Josh: ai compute to build a new model so we have an official deal there we have a 17 00:00:58,200 --> 00:01:02,200 Josh: new hardware device in the world of ai and we have just a bunch of cool space 18 00:01:02,200 --> 00:01:06,560 Josh: related stuff so you just let's get into the big news of the day which is open 19 00:01:06,560 --> 00:01:11,540 Josh: ai asking for a backstop that sounds scary walking That's exactly what happened. 20 00:01:11,680 --> 00:01:13,580 Josh: Because I feel like some things get lost in translation here. 21 00:01:13,660 --> 00:01:15,680 Josh: What exactly did they say and what does it mean for us? 22 00:01:15,800 --> 00:01:21,480 Ejaaz: Yeah. So let me set some context. Sarah Fryer is the current CFO at OpenAI. 23 00:01:21,660 --> 00:01:25,200 Ejaaz: And she did this interview with the Wall Street Journal, where obviously they 24 00:01:25,200 --> 00:01:28,700 Ejaaz: get into the economics of OpenAI's revenue model and how they're going to pay 25 00:01:28,700 --> 00:01:33,600 Ejaaz: for all these trillions of dollars worth of compute. And she had a few choice words. 26 00:01:34,280 --> 00:01:38,600 Ejaaz: The most choicest was around this backstop that you're talking about, 27 00:01:38,600 --> 00:01:41,600 Ejaaz: josh um let me let me kind of paint it out for you so 28 00:01:41,600 --> 00:01:45,300 Ejaaz: open ai has currently signed 1.4 trillion 29 00:01:45,300 --> 00:01:51,180 Ejaaz: dollars worth of compute deals this is open ai agreeing to pay compute providers 30 00:01:51,180 --> 00:01:56,180 Ejaaz: or chip providers or gpu providers like nvidia amd hundreds of billions of dollars 31 00:01:56,180 --> 00:02:00,960 Ejaaz: in order to buy their gpu so that they can train the next best ai model agi 32 00:02:00,960 --> 00:02:02,880 Ejaaz: there's one massive problem, 33 00:02:03,350 --> 00:02:06,290 Ejaaz: they don't have enough cash. They don't have enough money. In fact, 34 00:02:06,410 --> 00:02:08,810 Ejaaz: they are currently running at a loss. 35 00:02:09,190 --> 00:02:14,610 Ejaaz: There's no plan or near-term kind of goal that they can achieve to pay for this stuff. 36 00:02:14,870 --> 00:02:18,710 Ejaaz: So then the question becomes, what happens if they can't pay for this stuff? 37 00:02:18,790 --> 00:02:23,270 Ejaaz: What happens if Chachupiti stops becoming profitable in the future as they're 38 00:02:23,270 --> 00:02:24,850 Ejaaz: projected? What happens then? 39 00:02:25,350 --> 00:02:30,250 Ejaaz: And Sarah had one simple answer to that, which was the government can bail us out. 40 00:02:30,950 --> 00:02:36,730 Ejaaz: And so specifically what she described was an ecosystem where the government 41 00:02:36,730 --> 00:02:42,490 Ejaaz: basically pays OpenAI and gives them the money that they need to buy these GPUs 42 00:02:42,490 --> 00:02:44,410 Ejaaz: if they happen to default, 43 00:02:44,630 --> 00:02:47,530 Ejaaz: if they happen to not have enough money to pay for these things. 44 00:02:47,850 --> 00:02:53,170 Ejaaz: This, in turn, will give the private equity firms and the banks that are agreeing 45 00:02:53,170 --> 00:02:58,490 Ejaaz: to loan OpenAI money in the first place to buy these GPUs, the sanctity and peace of mind that 46 00:02:58,850 --> 00:03:03,610 Ejaaz: ah, I'm going to be okay. I'm going to get my money back if OpenAI doesn't deliver through. 47 00:03:04,050 --> 00:03:09,150 Ejaaz: The craziest part about this, Josh, is I feel like I've just been ricocheted 48 00:03:09,150 --> 00:03:11,370 Ejaaz: across the room for two weeks straight. 49 00:03:11,570 --> 00:03:16,590 Ejaaz: Because Sam Altman went from operating a non-profit to operating a non-profit 50 00:03:16,590 --> 00:03:22,250 Ejaaz: that is kind of a private company, where they kind of turned a part of their 51 00:03:22,250 --> 00:03:24,770 Ejaaz: company into a for-profit, so they're kind of like lying about it. 52 00:03:24,930 --> 00:03:28,850 Ejaaz: Then kind of breaking rumors around doing a $1 trillion IPO. 53 00:03:30,070 --> 00:03:34,550 Ejaaz: What is your immediate gut reaction to this? Is it lies? Is it real? What's happening? 54 00:03:34,590 --> 00:03:37,350 Josh: I just like, I keep repeating the words too big to fail in my head, 55 00:03:37,430 --> 00:03:38,670 Josh: too big to fail, too big to fail. 56 00:03:39,730 --> 00:03:43,870 Josh: It seems like that's what I want to get to. And I'm like, I'm of two minds of 57 00:03:43,870 --> 00:03:47,030 Josh: this. One is that, well, AI is a matter of national security. 58 00:03:47,210 --> 00:03:49,490 Josh: It is very important to get this right and to move as fast as possible. 59 00:03:49,690 --> 00:03:53,530 Josh: If that requires some government help, that probably make sense to an extent. 60 00:03:53,750 --> 00:03:57,850 Josh: And then the second thing is, well, if you are asking for government help, 61 00:03:58,150 --> 00:04:02,390 Josh: that's probably not a good thing in this instance. 62 00:04:02,630 --> 00:04:06,970 Josh: And government help really means taxpayer help. Like we fund this stuff. 63 00:04:07,130 --> 00:04:14,270 Josh: And to pay a backstop for Sam Waltman when we don't even get public stock exposure 64 00:04:14,270 --> 00:04:15,610 Josh: because it's still privately held. 65 00:04:15,750 --> 00:04:20,090 Josh: And another thing is she admitted here that they actually have little to no 66 00:04:20,090 --> 00:04:21,770 Josh: interest in IPO-ing anytime soon. 67 00:04:21,930 --> 00:04:25,870 Josh: So now there's no real trajectory for the public being able to own any upside 68 00:04:25,870 --> 00:04:27,710 Josh: and only participate in the downside. 69 00:04:28,170 --> 00:04:32,650 Josh: After they've been signaling that they are not for profit since inception, 70 00:04:32,650 --> 00:04:36,970 Josh: it's just like these really horrifically mixed signals with no clear intention. 71 00:04:36,970 --> 00:04:37,450 Ejaaz: It's exhausting. 72 00:04:38,050 --> 00:04:41,410 Josh: Um, it's a lot. Yeah. I guess the, the way that you described it being like 73 00:04:41,410 --> 00:04:43,890 Josh: ricocheted across a room, it like, it kind of feels that way. 74 00:04:44,030 --> 00:04:48,570 Josh: And it's a little disturbing, but I wonder if this is kind of at the core of 75 00:04:48,570 --> 00:04:52,330 Josh: what we've been seeing with this circular economic thing happening where opening, 76 00:04:52,470 --> 00:04:53,770 Josh: I sent to deal with Google with Microsoft. 77 00:04:53,790 --> 00:04:57,750 Josh: It's all very, it feels very incestual, but maybe it's because it really is 78 00:04:57,750 --> 00:04:59,950 Josh: a matter of national security and the government's just kind of allowing them 79 00:04:59,950 --> 00:05:03,910 Josh: to do a lot of things that wouldn't traditionally have been acceptable. 80 00:05:04,760 --> 00:05:07,840 Josh: But it looks like we have some qualifications, right? Yeah, I mean, 81 00:05:08,040 --> 00:05:09,340 Josh: maybe we should take this as a grain of salt. 82 00:05:10,000 --> 00:05:15,320 Ejaaz: Yeah, so a grain of salt being that Sarah Fry, the same CFO that kind of made 83 00:05:15,320 --> 00:05:20,760 Ejaaz: these claims, walked back her claims in the interview, specifying that OpenAI 84 00:05:20,760 --> 00:05:24,480 Ejaaz: is not seeking a government backstop for our infrastructure commitments. 85 00:05:24,480 --> 00:05:26,940 Ejaaz: And she goes on to give this official statement about, you know, 86 00:05:27,060 --> 00:05:30,920 Ejaaz: OpenAI, you know, being profitable and going to be able to pay their way through 87 00:05:30,920 --> 00:05:36,520 Ejaaz: this all. The issue that I have with this, Josh, is I don't think it's an honest statement. 88 00:05:36,920 --> 00:05:40,220 Ejaaz: Why don't I think it's an honest statement? Because her boss. 89 00:05:40,920 --> 00:05:45,920 Ejaaz: CEO Sam Altman, has claimed so many times in interviews and his own blog posts 90 00:05:45,920 --> 00:05:50,860 Ejaaz: that he's written that he has no issue asking the government for a bailout, for a backstop, 91 00:05:51,060 --> 00:05:56,600 Ejaaz: to help him kind of like cover his deficits and his debts if he's not able to pay for it. 92 00:05:56,720 --> 00:06:00,700 Ejaaz: I have an excerpt from an interview pulled up here where he goes, 93 00:06:00,840 --> 00:06:05,020 Ejaaz: at some level, when something gets sufficiently huge, whether or not they are 94 00:06:05,020 --> 00:06:08,740 Ejaaz: on paper, the federal government is kind of an insurer of last resort. 95 00:06:08,820 --> 00:06:12,960 Ejaaz: As we've seen in various financial crises and insurance companies screwing things 96 00:06:12,960 --> 00:06:14,740 Ejaaz: up, they basically cover us. 97 00:06:15,140 --> 00:06:19,620 Ejaaz: And so this is again, and this is the second time we've spoken about this this 98 00:06:19,620 --> 00:06:25,400 Ejaaz: week, about Sam Altman's character inconsistencies, where he says one thing and means another. 99 00:06:25,660 --> 00:06:29,020 Ejaaz: We did an episode earlier this week where we covered a 52-page deposition, 100 00:06:29,240 --> 00:06:35,400 Ejaaz: where his co-founder and former chief science officer, Ilya Sutskeva, 101 00:06:35,680 --> 00:06:40,380 Ejaaz: basically says that Sam was like incessantly lying and that's what led to his 102 00:06:40,380 --> 00:06:42,740 Ejaaz: ousting, his firing in November, 2023. 103 00:06:43,870 --> 00:06:46,850 Ejaaz: Fast forward to present day, it seems like Sam's still at it, Josh. 104 00:06:47,790 --> 00:06:51,630 Josh: There's a lot of character inconsistency. And what's certainly not helping the 105 00:06:51,630 --> 00:06:56,190 Josh: case is all of his co-founders from day one to up until a few years ago, 106 00:06:56,530 --> 00:07:00,110 Josh: they're all testifying against him. Elon strongly disagrees with Sam. 107 00:07:00,510 --> 00:07:05,810 Josh: Ilya and Mira, they strongly distrust Sam. So this is not, this is poor signal 108 00:07:05,810 --> 00:07:09,590 Josh: coming from the public perception, but also from people who know him personally, 109 00:07:09,730 --> 00:07:11,330 Josh: which is not really making a good case. 110 00:07:11,450 --> 00:07:14,450 Josh: So we'll see. We'll continue to monitor the situation. That's kind of where 111 00:07:14,450 --> 00:07:15,530 Josh: we're at with OpenAI News. 112 00:07:15,870 --> 00:07:19,390 Ejaaz: I just want to... Sorry, I think we need to just... 113 00:07:20,230 --> 00:07:23,630 Ejaaz: Look at the other side very quickly, Josh, and I'm curious of your take here 114 00:07:23,630 --> 00:07:25,490 Ejaaz: specifically. Yeah, let's get into the bubble. 115 00:07:25,890 --> 00:07:31,310 Ejaaz: So the take from this is, oh my God, we're doing circular investing. 116 00:07:31,650 --> 00:07:34,650 Ejaaz: This is a massive bubble. It's so obviously going to pop. 117 00:07:34,790 --> 00:07:38,870 Ejaaz: They're asking for a literal government bailout before the crash actually happens. 118 00:07:38,950 --> 00:07:42,030 Ejaaz: We've seen this with COVID. We've seen this in 2008. What are we doing here? 119 00:07:42,230 --> 00:07:46,710 Ejaaz: And the counter argument to that is, if you look at every other hyperscaler, 120 00:07:46,810 --> 00:07:48,850 Ejaaz: maybe not put open AI side for a second. 121 00:07:48,970 --> 00:07:52,210 Ejaaz: If you look at Meta, if you look at Google, if you look at Microsoft, 122 00:07:53,030 --> 00:07:55,590 Ejaaz: although they're spending hundreds of billions of dollars and committing to 123 00:07:55,590 --> 00:08:01,070 Ejaaz: do that over the next couple of years, they still haven't made a crazy enough 124 00:08:01,070 --> 00:08:05,390 Ejaaz: dent where you should be getting worried on their profit and loss sheet. 125 00:08:05,590 --> 00:08:09,350 Ejaaz: Remember, these companies all have other businesses that are massively profitable 126 00:08:09,350 --> 00:08:10,970 Ejaaz: and they're making tons and tons of money. 127 00:08:11,130 --> 00:08:14,210 Ejaaz: The money that they're investing in compute right now, technically, 128 00:08:14,450 --> 00:08:17,290 Ejaaz: if you weigh it up against that, isn't too crazy. 129 00:08:17,470 --> 00:08:21,450 Ejaaz: It is crazy. It is a bigger dent that we've seen them spend on anything else 130 00:08:21,450 --> 00:08:27,050 Ejaaz: over the last decade, but it's still not overcompensating for what they're earning right now. 131 00:08:27,230 --> 00:08:29,310 Ejaaz: And that's the only argument against it, which is like, hey, 132 00:08:29,510 --> 00:08:31,110 Ejaaz: we're seeing demand with our enterprise customers. 133 00:08:31,390 --> 00:08:34,010 Ejaaz: We're seeing demand with our retail customers. And so it makes sense for us 134 00:08:34,010 --> 00:08:36,930 Ejaaz: to invest in this compute. I don't know what you think about this. 135 00:08:37,250 --> 00:08:40,210 Josh: Yeah. Like in the case that AI scaling laws stop tomorrow, 136 00:08:40,490 --> 00:08:43,410 Josh: where suddenly we figure out oh oh no like this this isn't actually 137 00:08:43,410 --> 00:08:46,730 Josh: going to work spending more money the market gets harmed very 138 00:08:46,730 --> 00:08:52,290 Josh: badly but it's not catastrophic it is not a recession level bubble um in the 139 00:08:52,290 --> 00:08:57,110 Josh: case that this continues to prolong companies like open ai that don't have a 140 00:08:57,110 --> 00:09:01,770 Josh: cap table like companies like microsoft and google who exist outside of ai and 141 00:09:01,770 --> 00:09:04,690 Josh: are now using their balance sheet to pay this off i think that's probably when 142 00:09:04,690 --> 00:09:05,650 Josh: you start to see problems so 143 00:09:06,190 --> 00:09:08,510 Josh: I kind of agree with you in the sense that we're still good. 144 00:09:08,810 --> 00:09:13,890 Josh: Like things are still good. I'm not particularly concerned of a short-term bubble happening here. 145 00:09:14,030 --> 00:09:18,090 Ejaaz: I want to emphasize that like OpenAI isn't just kind of sitting on their hands 146 00:09:18,090 --> 00:09:20,510 Ejaaz: and not coming up with other ways to turn on revenue. 147 00:09:21,070 --> 00:09:25,770 Ejaaz: One other thing that she revealed in this interview, Josh, was I think it's 148 00:09:25,770 --> 00:09:27,170 Ejaaz: number six on the screen here. 149 00:09:27,610 --> 00:09:30,570 Ejaaz: Additionally, they will do creative commercial deals. 150 00:09:31,090 --> 00:09:35,590 Ejaaz: What they mean by that is if there's a company, say a pharmaceutical company 151 00:09:35,590 --> 00:09:41,050 Ejaaz: that uses ChatGPT and finds a cure for cancer using ChatGPT, 152 00:09:41,590 --> 00:09:45,590 Ejaaz: they're signing a deal with that pharmaceutical company such that they get a 153 00:09:45,590 --> 00:09:50,590 Ejaaz: percentage of profits from that drug that they create using their AI intelligence 154 00:09:50,590 --> 00:09:54,710 Ejaaz: that will occur for God knows how long after that. 155 00:09:55,070 --> 00:09:59,350 Ejaaz: And so taking a percentage of profit or revenue share from people who are using 156 00:09:59,350 --> 00:10:05,070 Ejaaz: it as a product to create other products, or if you're a company that sells products via ChatGPT, 157 00:10:05,230 --> 00:10:10,250 Ejaaz: Etsy is a common example that is live on chat gpt right now um they get a percentage 158 00:10:10,250 --> 00:10:14,190 Ejaaz: of profits um and then there's the obvious one which is chat gpt is going to 159 00:10:14,190 --> 00:10:18,290 Ejaaz: open air is just going to turn on ads and when they turn on ads who knows how 160 00:10:18,290 --> 00:10:21,230 Ejaaz: much money that's going to bring in so they are making efforts towards i don't 161 00:10:21,230 --> 00:10:22,730 Ejaaz: want this to be like a hey like 162 00:10:23,150 --> 00:10:28,070 Ejaaz: bad open ai thing but it's just unlikely given the amount that they've committed to spend. 163 00:10:28,070 --> 00:10:31,450 Josh: That first point you mentioned around like health breakthroughs 164 00:10:31,450 --> 00:10:34,030 Josh: through chat gpt getting a percentage that sounds like a train wreck waiting 165 00:10:34,030 --> 00:10:38,370 Josh: to happen that's a very messy monetization structure so a lot still to be evaluated 166 00:10:38,370 --> 00:10:42,390 Josh: but you just i want to talk about apple now we had an episode yesterday with 167 00:10:42,390 --> 00:10:45,730 Josh: apple um and google and how they kind of relate to each other particularly around 168 00:10:45,730 --> 00:10:50,530 Josh: a deal in which apple kind of sucks at ai they're really just not good at it and they need help. 169 00:10:50,790 --> 00:10:55,130 Josh: And here is Gemini coming to the rescue. We officially have a deal that is unofficially 170 00:10:55,130 --> 00:10:58,870 Josh: official, and it looks like they're going to be paying $1 billion a year, 171 00:10:59,090 --> 00:11:04,950 Josh: Apple, to Google in order to get a 1.2 trillion parameter Google Gemini model custom for Apple. 172 00:11:05,170 --> 00:11:07,230 Josh: This is a really big deal. Apple is struggling. 173 00:11:07,610 --> 00:11:12,110 Josh: Apple has not done anything in the world of AI. And suddenly they have this really powerful model. 174 00:11:12,250 --> 00:11:17,770 Josh: So this seems like it's going to be very important for the case of the bull case for apple at 175 00:11:17,770 --> 00:11:22,190 Ejaaz: Least so i'm happy about this uh for a few reasons normally i'd be laughing 176 00:11:22,190 --> 00:11:25,510 Ejaaz: right i'd be like haha this obviously apple's failed they're. 177 00:11:29,080 --> 00:11:33,160 Ejaaz: In Apple's sense, it's kind of a smart move. Think about it, right? 178 00:11:33,480 --> 00:11:37,620 Ejaaz: They haven't spent hundreds of billions of dollars trying to invest in GPUs 179 00:11:37,620 --> 00:11:39,440 Ejaaz: and train a complex AI model. 180 00:11:39,680 --> 00:11:43,260 Ejaaz: They haven't taken on any of that risk. They just tap Google on the shoulder, 181 00:11:43,300 --> 00:11:44,600 Ejaaz: who's done all the hard work and say, 182 00:11:45,060 --> 00:11:48,900 Ejaaz: yo, are you down if I pay you a billion dollars per year and you make our own 183 00:11:48,900 --> 00:11:54,260 Ejaaz: custom version of an Apple AI model that I can plug into Siri and will run on 184 00:11:54,260 --> 00:11:55,600 Ejaaz: our private cloud instance? 185 00:11:55,860 --> 00:11:57,980 Ejaaz: So, you know, Google won't necessarily get access to all of it, 186 00:11:58,060 --> 00:11:59,580 Ejaaz: but they just get the payment every year. 187 00:11:59,780 --> 00:12:03,900 Ejaaz: That sounds like a pretty sweet deal. The other thing I like about this is this 188 00:12:03,900 --> 00:12:06,100 Ejaaz: isn't just any kind of like model. 189 00:12:06,220 --> 00:12:08,580 Ejaaz: It's a 1.2 trillion parameter model. 190 00:12:09,000 --> 00:12:12,440 Ejaaz: That is like up there with like one of the biggest models that would be out there. 191 00:12:12,660 --> 00:12:17,200 Ejaaz: And to combine that with the kind of personalization that I'm presuming Apple 192 00:12:17,200 --> 00:12:22,600 Ejaaz: is going to integrate into Siri and in the consumer experience with using text 193 00:12:22,600 --> 00:12:25,540 Ejaaz: and other apps on the phone, that's pretty attractive to me. 194 00:12:25,540 --> 00:12:29,100 Ejaaz: The other thing that I thought was super cool on the Google side here, Josh, 195 00:12:29,260 --> 00:12:35,980 Ejaaz: is to be able to run a 1.2 trillion parameter model at an economically viable 196 00:12:35,980 --> 00:12:39,900 Ejaaz: cost, aka they're making money from that, just goes to show that there's some 197 00:12:39,900 --> 00:12:43,280 Ejaaz: pretty crazy engineering that Google has achieved. 198 00:12:43,460 --> 00:12:47,740 Ejaaz: This indirectly tells me that they have absolutely nailed their chip design 199 00:12:47,740 --> 00:12:51,560 Ejaaz: and their TPU architecture to be able to pull this off. Just super cool. 200 00:12:51,560 --> 00:12:53,080 Josh: There's a lot more info if you 201 00:12:53,080 --> 00:12:55,260 Josh: want to find out about Apple on our episode that we released yesterday. 202 00:12:55,500 --> 00:13:00,400 Josh: And also I am publishing an essay in the newsletter today when you're watching 203 00:13:00,400 --> 00:13:03,580 Josh: this all about the economics and why this makes sense for Apple to do. 204 00:13:03,720 --> 00:13:06,640 Josh: So if you're interested in hearing more of these takes, like more thoughtful 205 00:13:06,640 --> 00:13:08,260 Josh: takes, check it out on the newsletter. 206 00:13:08,540 --> 00:13:11,200 Josh: I'm like very proud of this article. I think it'll do really well. 207 00:13:11,560 --> 00:13:16,540 Josh: It's just really fascinating to see the Apple strategy kind of accidentally step into this 208 00:13:17,080 --> 00:13:19,960 Josh: amazing situation for them um where it was 209 00:13:19,960 --> 00:13:22,780 Josh: certainly not by design but they somehow managed to put themselves in a 210 00:13:22,780 --> 00:13:26,020 Josh: really good place but on the topic of google i also want to talk about the new 211 00:13:26,020 --> 00:13:30,140 Josh: hardware that they just announced which is their new ironwood tpus now again 212 00:13:30,140 --> 00:13:33,680 Josh: in yesterday's episode this is a good one we talked about what a tpu was and 213 00:13:33,680 --> 00:13:38,360 Josh: how it relates to a gpu and today in some new news we got new tpus ej so can 214 00:13:38,360 --> 00:13:41,560 Josh: you walk us through what these ironwoods are what they do why they're impressive so 215 00:13:41,560 --> 00:13:46,080 Ejaaz: Ironwood is google's latest tpu tpu stands for tensor processing unit. 216 00:13:46,580 --> 00:13:51,680 Ejaaz: All you need to know is that the TPU of Google is the equivalent of the GPU 217 00:13:51,680 --> 00:13:54,140 Ejaaz: from NVIDIA, but with some additional perks. 218 00:13:54,280 --> 00:14:01,200 Ejaaz: It is more specialized and custom fit towards Google software and AppSuite, right? 219 00:14:01,240 --> 00:14:03,720 Ejaaz: It's the thing that has powered and trained all their models. 220 00:14:03,900 --> 00:14:08,120 Ejaaz: They've never actually relied on NVIDIA at all to kind of train and inference 221 00:14:08,120 --> 00:14:10,540 Ejaaz: a bunch of their models. They've been kind of like this lone entity. 222 00:14:11,020 --> 00:14:14,020 Ejaaz: And why this is so cool, and we explained this on yesterday's episode, 223 00:14:14,140 --> 00:14:14,960 Ejaaz: which you should check out, 224 00:14:15,420 --> 00:14:19,860 Ejaaz: is Google's been super independent and they've been able to make several breakthroughs, 225 00:14:19,920 --> 00:14:25,640 Ejaaz: which have allowed them to train the same type of models that OpenAI and Microsoft produce, 226 00:14:25,860 --> 00:14:31,780 Ejaaz: but much cheaper or cost efficient and can scale massively with a greater number 227 00:14:31,780 --> 00:14:32,780 Ejaaz: of models that they build. 228 00:14:33,020 --> 00:14:36,320 Ejaaz: And now they've released this new model called Ironwood, which is basically 229 00:14:36,320 --> 00:14:38,560 Ejaaz: the next generation of their TPU. 230 00:14:38,820 --> 00:14:44,940 Ejaaz: It is four times faster than the prior version, and it can basically clamp together 231 00:14:44,940 --> 00:14:49,080 Ejaaz: as one singular stack in a much more feasible way, which means that training 232 00:14:49,080 --> 00:14:51,820 Ejaaz: larger models at scale is going to be much easier. 233 00:14:52,540 --> 00:14:57,280 Ejaaz: But the biggest news about this for me, Josh, is they're going to start selling 234 00:14:57,280 --> 00:15:02,720 Ejaaz: these TPUs and making it more accessible for anyone else to buy their TPUs and 235 00:15:02,720 --> 00:15:05,260 Ejaaz: train or influence their own AI models. 236 00:15:05,690 --> 00:15:11,670 Ejaaz: Why this is such big news is this means they're formally stepping into the ring to compete with NVIDIA. 237 00:15:11,830 --> 00:15:16,210 Ejaaz: Now, don't get me wrong. They're not doing this at the scale that NVIDIA is currently doing it. 238 00:15:16,470 --> 00:15:19,770 Ejaaz: But currently, there has been no feasible challenger to NVIDIA. 239 00:15:19,970 --> 00:15:23,970 Ejaaz: And now you have Google entering the ring, which has a lot of distribution and technical expertise. 240 00:15:24,830 --> 00:15:29,830 Ejaaz: It's notable and probably a hint that the Google market cap should be much, much higher. 241 00:15:30,130 --> 00:15:33,530 Josh: Yeah, I wouldn't say no one is competing with NVIDIA. There are AMD chips. 242 00:15:33,630 --> 00:15:36,890 Josh: China's creating their own alternatives. So I'd say it's at that level where 243 00:15:36,890 --> 00:15:40,290 Josh: it's trying to compete, but there is going to be a very steep mountain in order 244 00:15:40,290 --> 00:15:41,830 Josh: to get there, to become a real NVIDIA competitor. 245 00:15:42,010 --> 00:15:45,910 Josh: In other news, we got a new AI hardware device this week, and it goes by the 246 00:15:45,910 --> 00:15:49,270 Josh: name of Sandbar. And it comes in the form factor of a ring. 247 00:15:49,470 --> 00:15:53,830 Josh: Now, I have an Oura ring on it. I love my Oura ring. It's a great non-intrusive sleep tracker. 248 00:15:54,230 --> 00:15:58,610 Josh: This is a totally new take on a ring because it gets into these things that we call edge nodes. 249 00:15:58,610 --> 00:16:01,330 Josh: And when you deal with ai systems there there are 250 00:16:01,330 --> 00:16:04,550 Josh: sensors that can then send requests back to the 251 00:16:04,550 --> 00:16:07,790 Josh: actual ai system and this is a new sensor so what 252 00:16:07,790 --> 00:16:11,770 Josh: we're seeing on screen is a promo video of this person who has a ring you press 253 00:16:11,770 --> 00:16:14,570 Josh: a button on the ring and you can speak into a microphone now this microphone 254 00:16:14,570 --> 00:16:19,710 Josh: acts as an interface between yourself and the ai system and you could ask it 255 00:16:19,710 --> 00:16:25,370 Josh: to do things like remind you to query questions against it to record conversations and it's this really 256 00:16:25,870 --> 00:16:30,430 Josh: unique and I guess somewhat novel form factor in the world of AI hardware devices. 257 00:16:30,570 --> 00:16:32,250 Josh: This was interesting to me, Ejaz, at least because 258 00:16:32,770 --> 00:16:38,090 Josh: I'm so fascinated about what OpenAI is going to make next year with Johnny Ive in terms of AI hardware. 259 00:16:38,370 --> 00:16:42,190 Josh: And this is an interesting experiment to kind of see how the ring form factor would work. 260 00:16:42,310 --> 00:16:47,150 Josh: So the way this kind of exists is it's a microphone that has passive audio back 261 00:16:47,150 --> 00:16:49,490 Josh: to your earbuds, or I assume back to your phone. 262 00:16:49,730 --> 00:16:55,790 Josh: And it's this really fun and somewhat intuitive way of using AI without AI getting in the way. 263 00:16:55,950 --> 00:16:59,030 Josh: So I think a lot of the things that we're going to start to see in this 264 00:16:59,030 --> 00:17:02,090 Josh: removal of the smartphone is this the suite of 265 00:17:02,090 --> 00:17:05,230 Josh: ambient devices where you can just kind of engage with ai wherever you 266 00:17:05,230 --> 00:17:08,070 Josh: are at any time and a ring is a really neat form factor for 267 00:17:08,070 --> 00:17:10,670 Josh: this because it's it doesn't really get in the way it's just kind of 268 00:17:10,670 --> 00:17:14,670 Josh: always there if you want to engage with it you summon it if you don't you don't 269 00:17:14,670 --> 00:17:18,990 Josh: even think about it and i think this is this is an interesting experiment and 270 00:17:18,990 --> 00:17:22,190 Josh: it's something i'd kind of want to try this not that i think it's a successful 271 00:17:22,190 --> 00:17:23,910 Josh: product but i think it's an it's 272 00:17:23,910 --> 00:17:28,190 Josh: an interesting take on what the future ai devices could look like yeah 273 00:17:28,190 --> 00:17:30,910 Ejaaz: Well where my mind immediately leaps when i 274 00:17:30,910 --> 00:17:33,710 Ejaaz: i look at this is um her holding up 275 00:17:33,710 --> 00:17:36,550 Ejaaz: her hand to speak into the ring it either looks like 276 00:17:36,550 --> 00:17:39,370 Ejaaz: she's about to cough or like you know in the movies where 277 00:17:39,370 --> 00:17:42,150 Ejaaz: the the bodyguards are like touching their ear when they need to speak 278 00:17:42,150 --> 00:17:44,990 Ejaaz: into the secret service it kind 279 00:17:44,990 --> 00:17:48,710 Ejaaz: of seems like something like that so habitually i'm kind of curious as to how 280 00:17:48,710 --> 00:17:51,730 Ejaaz: this kind of integrates into society but i agree with you i think like the ring 281 00:17:51,730 --> 00:17:56,250 Ejaaz: is super subtle and kind of non-obtrusive and it makes it super convenient to 282 00:17:56,250 --> 00:17:59,790 Ejaaz: kind of engage with this technology without needing to stare at another screen 283 00:17:59,790 --> 00:18:03,490 Ejaaz: and whilst maintaining the ability to kind of interact with real life. 284 00:18:03,790 --> 00:18:07,370 Josh: Yeah, this led me down a rabbit hole because we were on the topic of Apple and 285 00:18:07,370 --> 00:18:10,450 Josh: it really, I'd love for Apple to start doing things like this. 286 00:18:10,830 --> 00:18:14,470 Josh: Like for example, Apple has a hundred billion dollars in cash. They go acquire Aura. 287 00:18:14,850 --> 00:18:19,150 Josh: Now they get the custom Gemini model from Google. Suddenly they have AI. They ship this 288 00:18:19,530 --> 00:18:23,430 Josh: this ai ring that is compatible with these new gemini models they ship a new 289 00:18:23,430 --> 00:18:26,730 Josh: set of airpods that have visual sensors on them so you could collect data from 290 00:18:26,730 --> 00:18:30,610 Josh: the outside world and you start to get this suite of devices that isn't an iphone 291 00:18:30,610 --> 00:18:35,490 Josh: but is increasingly capable and more powerful and approaching what we can do 292 00:18:35,490 --> 00:18:38,830 Josh: with an iphone so i hope this is a trend that we see where the next 293 00:18:39,290 --> 00:18:43,210 Josh: iphone level device isn't a device it's a suite of devices maybe the ring is 294 00:18:43,210 --> 00:18:46,350 Josh: one of them maybe it's not but it's an interesting experiment to see what it 295 00:18:46,350 --> 00:18:48,770 Josh: could look like if that became the case I 296 00:18:48,770 --> 00:18:53,210 Ejaaz: Want to talk about the dark horse of the AI race, which is Anthropic. 297 00:18:53,430 --> 00:18:57,530 Ejaaz: Now, I'm going to hold my hands up here, Josh. I have given Anthropic a lot 298 00:18:57,530 --> 00:19:00,790 Ejaaz: of flack, and I've kind of called them the narc AI model. 299 00:19:00,950 --> 00:19:04,170 Ejaaz: They follow the rules, they go to the government, they run to the government 300 00:19:04,170 --> 00:19:06,330 Ejaaz: and say, hey, can you give us this deal or whatever? 301 00:19:06,550 --> 00:19:11,250 Ejaaz: And I've kind of looked down on Claude since they kind of maintained their parity 302 00:19:11,250 --> 00:19:14,850 Ejaaz: at the coding agent level. So I was kind of like, whatever, Why would I use Claude? 303 00:19:15,290 --> 00:19:20,930 Ejaaz: Turns out I was very wrong. So the information leaked a report on projected 304 00:19:20,930 --> 00:19:26,530 Ejaaz: revenue for Anthropic, and it basically puts them at the same level as OpenAI, 305 00:19:26,710 --> 00:19:28,750 Ejaaz: in some cases, better, Josh. 306 00:19:29,230 --> 00:19:34,650 Ejaaz: So the major takeaway from this is they're projecting $70 billion worth of revenue 307 00:19:34,650 --> 00:19:38,610 Ejaaz: by 2028 and a $400 billion valuation. 308 00:19:38,910 --> 00:19:42,310 Ejaaz: Bear in mind that OpenAI is currently valued at, I think, $500 billion. 309 00:19:42,730 --> 00:19:48,730 Ejaaz: So to make that leap from where they are currently, which I think is $200 billion, is a crazy jump. 310 00:19:49,010 --> 00:19:55,470 Ejaaz: But number two, this would assume they then become profitable way earlier than OpenAI. 311 00:19:55,670 --> 00:20:00,190 Ejaaz: OpenAI at the same time, 2028, will be making technically more money than Anthropic, 312 00:20:00,550 --> 00:20:02,150 Ejaaz: but won't be profitable. 313 00:20:02,590 --> 00:20:05,970 Ejaaz: And so the immediate question that I asked myself was, well, 314 00:20:06,210 --> 00:20:08,010 Ejaaz: how are they planning to do this? 315 00:20:08,110 --> 00:20:11,870 Ejaaz: So far, they're losing on the retail sales. So Like, do they have a Hail Mary? 316 00:20:12,170 --> 00:20:16,470 Ejaaz: And the answer I connected to the dots, Josh, comes right here. 317 00:20:16,960 --> 00:20:24,020 Ejaaz: In the enterprise AI market share, where Anthropic has sneakily surpassed OpenAI. 318 00:20:24,240 --> 00:20:26,520 Ejaaz: They currently command, I think this chart is a little outdated, 319 00:20:26,680 --> 00:20:31,240 Ejaaz: but they currently command around 25% of the enterprise API share. 320 00:20:31,400 --> 00:20:36,420 Ejaaz: And why this is super important is, although there may be fewer enterprise customers 321 00:20:36,420 --> 00:20:41,900 Ejaaz: in terms of numbers, like so, you know, OpenAI has 800 million weekly active users. 322 00:20:42,080 --> 00:20:46,280 Ejaaz: Anthropic might have, I don't know, a couple hundred thousand enterprise users. 323 00:20:46,280 --> 00:20:51,560 Ejaaz: Each enterprise user pays way, way, way more than the average retail user. 324 00:20:51,740 --> 00:20:54,060 Ejaaz: And so it's just something that I didn't see, Josh. 325 00:20:54,340 --> 00:20:57,080 Ejaaz: They're like engaging with a lot of enterprises behind the scenes. 326 00:20:57,220 --> 00:21:01,380 Ejaaz: They're signing these multi-billion dollar contracts and they're actually like 327 00:21:01,380 --> 00:21:06,460 Ejaaz: translating these contracts into useful products that these businesses are using 328 00:21:06,460 --> 00:21:09,580 Ejaaz: behind the scenes. How we prove that, I don't know. 329 00:21:09,800 --> 00:21:13,500 Ejaaz: Maybe it's like the economic GDP over time from a bunch of these different companies 330 00:21:13,500 --> 00:21:15,680 Ejaaz: that they've signed deals, but I thought this was cool to point out. 331 00:21:16,060 --> 00:21:19,720 Josh: Yeah, this to me intuitively makes sense. It's like when you think of coding, 332 00:21:19,880 --> 00:21:24,020 Josh: you think of Claude. And I think that's kind of the universal truth amongst corporations. 333 00:21:24,520 --> 00:21:28,640 Josh: And when you want a bot to write code for you, you are using Claude. 334 00:21:28,740 --> 00:21:32,880 Josh: And if you're pinging an API that is writing code for you, odds are you're using Claude. 335 00:21:32,980 --> 00:21:39,700 Josh: So while the retail general public facing sentiment isn't that optimistic around Claude and Anthropic, 336 00:21:39,900 --> 00:21:43,700 Josh: because it's just not as useful as ChatShapute or Gemini, The reality is that 337 00:21:43,700 --> 00:21:46,560 Josh: if you're writing code and if you are a company that wants a model that writes 338 00:21:46,560 --> 00:21:49,800 Josh: fantastic code, you are using Claude and Anthropic. 339 00:21:49,980 --> 00:21:52,900 Josh: And Anthropic is just collecting a lot of the upside without a lot of the public 340 00:21:52,900 --> 00:21:56,480 Josh: facing noise because those are just private entities. They're just swiping their 341 00:21:56,480 --> 00:22:00,140 Josh: credit cards and they're getting all their tokens and they're just happy and they're on their way. 342 00:22:00,320 --> 00:22:02,560 Josh: So, yeah, I mean, I'm happy for Anthropic. 343 00:22:03,360 --> 00:22:07,320 Josh: I hope this is durable. I like the fact that we're starting to see each of these 344 00:22:07,320 --> 00:22:11,120 Josh: companies kind of slot themselves into a portion of the market. 345 00:22:11,280 --> 00:22:14,580 Josh: So I like that Anthropic is just working on code. I think that's great. 346 00:22:14,900 --> 00:22:17,620 Josh: Don't try to be the best at everything. Try to make the best coding model and 347 00:22:17,620 --> 00:22:20,360 Josh: look how much money you could print from it. This is a positive sum game. 348 00:22:20,500 --> 00:22:22,240 Josh: The pie is continuing to grow so quickly. 349 00:22:22,420 --> 00:22:25,560 Josh: So if you could just own a small corner of it, like Anthropic is doing, 350 00:22:26,040 --> 00:22:27,400 Josh: all the power to them. Keep it going. 351 00:22:27,940 --> 00:22:33,980 Josh: Someone who's not faring as well as Anthropic is our good friends over at Meta and Mr. Zuckerberg. 352 00:22:34,280 --> 00:22:39,040 Josh: They got absolutely crushed this week after earnings for a series of reasons. 353 00:22:39,480 --> 00:22:44,220 Josh: This doesn't really come as a shock. I think we've been pretty bearish on Meta as a whole. 354 00:22:44,740 --> 00:22:49,300 Josh: After that whole glasses debacle, I became increasingly bearish on Meta as a 355 00:22:49,300 --> 00:22:51,700 Josh: company and their ability to execute in this world of AI. 356 00:22:52,180 --> 00:22:55,000 Josh: But Ejaz, do you have any takes on what happened after the Meta earnings report? 357 00:22:55,400 --> 00:23:01,520 Ejaaz: I do. So we put out an episode, I think two weeks ago, on the Meta bull case. 358 00:23:01,620 --> 00:23:04,360 Ejaaz: And so the question a lot of you might have is like, you know, 359 00:23:04,440 --> 00:23:05,980 Ejaaz: do you still maintain some of that? 360 00:23:06,240 --> 00:23:09,300 Ejaaz: The short answer is yes, but over a longer time period. 361 00:23:09,860 --> 00:23:16,140 Ejaaz: Like here's the facts. So earnings came out for Meta last week and it wasn't 362 00:23:16,140 --> 00:23:19,100 Ejaaz: as great as they'd hoped for one particular reason. 363 00:23:19,480 --> 00:23:23,940 Ejaaz: Their spend on AI was ludicrous for the quarter. 364 00:23:24,100 --> 00:23:26,980 Josh: The company spending tens of billions on employees is overspending. 365 00:23:27,380 --> 00:23:29,020 Josh: I would have never guessed. 366 00:23:29,020 --> 00:23:32,600 Ejaaz: Exactly. So they had spent billions and billions of dollars, 367 00:23:32,600 --> 00:23:36,400 Ejaaz: in some cases, in crazy ways, just to hire a couple of people, 368 00:23:36,560 --> 00:23:40,540 Ejaaz: but also to pay for CapEx investments for their Hyperion data center, 369 00:23:40,540 --> 00:23:44,140 Ejaaz: to invest in different apps being built, to fire a bunch of people, 370 00:23:44,380 --> 00:23:48,560 Ejaaz: just ludicrous amounts being spent. And I think that... 371 00:23:48,750 --> 00:23:53,730 Ejaaz: Meta shareholders had a bit of PTSD from the Metaverse days in 2022, 372 00:23:53,730 --> 00:23:58,490 Ejaaz: where they renamed their entire company from Facebook to Meta based on this 373 00:23:58,490 --> 00:24:00,750 Ejaaz: Metaverse theory and ended up not panning out, right? 374 00:24:01,270 --> 00:24:04,730 Ejaaz: NFTs weren't a thing. And so they have a bit of PTSD where they're seeing Zuck 375 00:24:04,730 --> 00:24:06,670 Ejaaz: spending all this money, but no real ROI. 376 00:24:07,430 --> 00:24:12,510 Ejaaz: One clear example is they launched an AI assistant and no one really uses it. 377 00:24:12,570 --> 00:24:15,430 Ejaaz: No one's really on Facebook and it's not really integrated well into their existing 378 00:24:15,430 --> 00:24:17,170 Ejaaz: products. People just use ChatGPT. 379 00:24:17,410 --> 00:24:22,410 Ejaaz: They then launched a Sora competitor called the MetaVibes app. 380 00:24:22,570 --> 00:24:25,130 Ejaaz: No one uses that either. Do you remember that, right? 381 00:24:25,470 --> 00:24:28,830 Ejaaz: And then they launched, they pioneered, they said, listen, we're going to stop 382 00:24:28,830 --> 00:24:30,950 Ejaaz: copying people. We're going to do our own thing. And they launched their own 383 00:24:30,950 --> 00:24:32,570 Ejaaz: hardware device, which is the AI glasses. 384 00:24:32,970 --> 00:24:37,790 Ejaaz: The reception and feedback from their diehard fans was the worst they've ever 385 00:24:37,790 --> 00:24:39,310 Ejaaz: heard it. They hated the entire experience. 386 00:24:39,410 --> 00:24:43,010 Ejaaz: They think it is a lesser product than anything else on the market. 387 00:24:43,130 --> 00:24:45,110 Ejaaz: So for all of these reasons and much more. 388 00:24:45,290 --> 00:24:49,330 Ejaaz: People just don't have faith in Zuck's ability to spend and deliver on this. 389 00:24:49,410 --> 00:24:50,970 Ejaaz: And it's reflecting in the share price. 390 00:24:51,090 --> 00:24:54,930 Ejaaz: They lost $250 billion in 24 hours, Josh, on market over. 391 00:24:55,310 --> 00:24:59,530 Ejaaz: Just completely insane, down 15%. So now going on to the bull case, 392 00:24:59,750 --> 00:25:01,650 Ejaaz: I do think they picked themselves out. 393 00:25:01,750 --> 00:25:04,590 Ejaaz: I actually think Meta is probably like a really good buy at this point. 394 00:25:04,970 --> 00:25:08,530 Ejaaz: And for one solid reason, which is I don't think Zuck wants to lose this race. 395 00:25:08,690 --> 00:25:13,110 Ejaaz: And he's willing to figure any and all out to get himself to a point. 396 00:25:13,710 --> 00:25:18,110 Ejaaz: Is he able to produce like a bunch of really cool apps that leverage kind of 397 00:25:18,110 --> 00:25:21,010 Ejaaz: like the distribution that he has using AI, like Google has, 398 00:25:21,310 --> 00:25:23,910 Ejaaz: like OpenAI has? I don't know. TBD. 399 00:25:24,610 --> 00:25:26,030 Ejaaz: But yeah, Josh, do you have any thoughts? 400 00:25:26,410 --> 00:25:28,670 Josh: There's this interesting phenomenon happening where 401 00:25:29,440 --> 00:25:33,120 Josh: I feel absolutely zero inclination to use any of Meta's products. 402 00:25:33,240 --> 00:25:35,840 Josh: And that doesn't exist with any other company. 403 00:25:35,980 --> 00:25:38,100 Josh: Like I've experimented with pretty much everything. 404 00:25:38,620 --> 00:25:44,440 Josh: There is not a single part of any of Meta's AI product stack that I'm remotely interested in. 405 00:25:44,680 --> 00:25:48,540 Josh: In fact, the only touch point I have with Meta is Instagram because Facebook 406 00:25:48,540 --> 00:25:52,060 Josh: is such a disaster and it's cluttered and it just doesn't, I haven't used it in years. 407 00:25:52,280 --> 00:25:59,040 Josh: So the Meta ecosystem as a whole is not interesting. the meta hardware delivery 408 00:25:59,040 --> 00:26:03,860 Josh: is horrific at best like ijaz you you wanted the glasses you ordered the glasses 409 00:26:03,860 --> 00:26:05,460 Josh: you still don't have the glass dude 410 00:26:05,460 --> 00:26:07,960 Ejaaz: I couldn't order the glasses i couldn't walk into the. 411 00:26:07,960 --> 00:26:10,820 Josh: Store yeah so so that's my point is to have 412 00:26:10,820 --> 00:26:13,740 Josh: a flagship product like that um be that 413 00:26:13,740 --> 00:26:16,540 Josh: despicable it's it's really it's it's high signal that there is 414 00:26:16,540 --> 00:26:19,880 Josh: some sort of lack of care in terms of what's being delivered and 415 00:26:19,880 --> 00:26:22,580 Josh: perhaps move fast and break things worked early on in the days of 416 00:26:22,580 --> 00:26:25,500 Josh: facebook but in a fully formed meta that is 417 00:26:25,500 --> 00:26:28,520 Josh: a huge behemoth now that doesn't work as 418 00:26:28,520 --> 00:26:32,620 Josh: well because the costs are so high i don't like 419 00:26:32,620 --> 00:26:35,360 Josh: anything about it the hardware even if 420 00:26:35,360 --> 00:26:40,600 Josh: it was exceptional even if meta released true augmented reality glasses today 421 00:26:40,600 --> 00:26:45,940 Josh: and they locked me into the facebook meta ecosystem i'm not a user so they could 422 00:26:45,940 --> 00:26:48,700 Josh: have the best product in the world the product that they hope to release five 423 00:26:48,700 --> 00:26:51,880 Josh: years from now even if they released it today i'm not a user because i don't 424 00:26:51,880 --> 00:26:54,820 Josh: care for that ecosystem and they don't care to unlock the ecosystem. 425 00:26:55,140 --> 00:26:57,700 Josh: So I think they have a lot of hard problems to solve. 426 00:26:58,060 --> 00:27:01,320 Josh: One is actually building a product people like in the hardware world. 427 00:27:01,460 --> 00:27:04,460 Josh: One is building a software stack that people like, and one is shifting these 428 00:27:04,460 --> 00:27:08,100 Josh: billions of users that they have over to something more meaningful than a social 429 00:27:08,100 --> 00:27:09,820 Josh: media feed. And maybe they don't do it. 430 00:27:09,920 --> 00:27:13,140 Josh: I don't know how they monetize in the case that they don't, but there's a lot 431 00:27:13,140 --> 00:27:17,460 Josh: of questions that need to be solved from Zuck and the Facebook team that remain to be 432 00:27:18,300 --> 00:27:18,840 Josh: unknown. 433 00:27:19,320 --> 00:27:25,180 Ejaaz: Josh, are you, is it hot where you are? You look to be, you're sweating a little bit. 434 00:27:26,340 --> 00:27:28,800 Ejaaz: Like, very, very interesting, because 435 00:27:28,800 --> 00:27:34,700 Ejaaz: in this next and final topic of the episode, I am pleased to announce, 436 00:27:35,660 --> 00:27:39,500 Ejaaz: Google is launching GPUs into space. 437 00:27:39,500 --> 00:27:42,080 Josh: I feel like I need boxing gloves every time I bring this up, 438 00:27:42,160 --> 00:27:48,480 Ejaaz: My God. To create an AI data center in space, which Josh hates so viscerally. 439 00:27:48,700 --> 00:27:51,880 Ejaaz: But one of the richest mans in the world is going to give you his argument as 440 00:27:51,880 --> 00:27:54,700 Ejaaz: to why it's an important thing to explore. 441 00:27:55,120 --> 00:28:00,980 Ejaaz: So termed Project Suncatcher, a moonshot attempt to launch GPUs into space by 442 00:28:00,980 --> 00:28:04,200 Ejaaz: Google to harness the power of the sun, 443 00:28:04,540 --> 00:28:10,620 Ejaaz: solar energy, which equates to 100 trillion times humanity's total electricity production. 444 00:28:10,920 --> 00:28:14,880 Ejaaz: These are Sundar Pichai's words, the CEO of Google, not mine, right? 445 00:28:15,120 --> 00:28:18,700 Ejaaz: He then goes on to explain how he would attempt to do that. 446 00:28:18,920 --> 00:28:21,480 Ejaaz: You know, he says like, listen, I know that there are problems in space. 447 00:28:21,620 --> 00:28:23,320 Ejaaz: There's radiation, but we're working on that. 448 00:28:23,420 --> 00:28:27,000 Ejaaz: We're running trials and tests right now, which actually proves that our TPUs 449 00:28:27,000 --> 00:28:28,340 Ejaaz: can survive in that radiation. 450 00:28:28,500 --> 00:28:31,760 Ejaaz: Okay, checkpoint number one. But of course, the question remains, 451 00:28:31,920 --> 00:28:36,660 Ejaaz: how on earth are you going to harness the energy that comes from the sun in 452 00:28:36,660 --> 00:28:38,720 Ejaaz: an efficient manner? Maybe that's easy to solve. 453 00:28:38,880 --> 00:28:44,040 Ejaaz: And then the obvious one is, it's so expensive to send stuff into space. 454 00:28:44,160 --> 00:28:48,740 Ejaaz: How are you going to pay for that? And actually, if you dig into his report and his announcement, 455 00:28:48,940 --> 00:28:53,320 Ejaaz: he argues and he makes the point that like, you know, if you extrapolate the 456 00:28:53,320 --> 00:28:57,580 Ejaaz: cost of space going forwards, it should end up being equivalently pretty cheap 457 00:28:57,580 --> 00:29:01,540 Ejaaz: for us to send GPUs where it makes sense to create a data center in space. 458 00:29:01,800 --> 00:29:06,660 Ejaaz: I'm going to pause there before I start gloating. But Josh, any feedback on this? 459 00:29:07,060 --> 00:29:10,380 Josh: OK, one thing that I do like is they're readjusting these timelines here. 460 00:29:10,520 --> 00:29:12,260 Josh: These timelines are starting to feel a little more realistic. 461 00:29:12,580 --> 00:29:14,760 Josh: Launch two prototypes by 2027. 462 00:29:15,600 --> 00:29:19,060 Josh: So we're pre-prototype now. This makes me happier. 463 00:29:19,240 --> 00:29:23,220 Josh: We are going to launch prototypes in two years. We are hopeful that SpaceX and 464 00:29:23,220 --> 00:29:26,820 Josh: Starship will be able to get the cost per kilogram to orbit down low enough where it makes sense. 465 00:29:27,280 --> 00:29:31,320 Josh: Fine. Okay. If you want to experiment with these moonshots and you want to take your time with it, 466 00:29:31,690 --> 00:29:34,450 Josh: all right i guess i give up like i'm i i got nothing else 467 00:29:34,450 --> 00:29:37,270 Josh: to say this seems like do it like you 468 00:29:37,270 --> 00:29:40,070 Josh: got the billions of dollars you put it into r&d you do it 469 00:29:40,070 --> 00:29:42,910 Josh: um i think it's like go go to 470 00:29:42,910 --> 00:29:45,690 Josh: space i don't care i'm just done i'm done fighting about this i got nothing 471 00:29:45,690 --> 00:29:48,770 Josh: else and also i'm running out of out of arguments because he is 472 00:29:48,770 --> 00:29:52,110 Josh: like methodically um removing the constraints by 473 00:29:52,110 --> 00:29:54,990 Josh: um i guess just just taking more time and assuming 474 00:29:54,990 --> 00:29:57,710 Josh: the unit economics makes sense the problem with doing it right now 475 00:29:57,710 --> 00:30:01,190 Josh: is the cost per kilogram to orbit is so unnecessarily high um 476 00:30:01,190 --> 00:30:04,430 Josh: and the technology is so difficult to prove as prototypes 477 00:30:04,430 --> 00:30:07,270 Josh: that it doesn't seem worthwhile i hope that 478 00:30:07,270 --> 00:30:10,490 Josh: they can figure it out i guess is what i have to say but anyway that that's 479 00:30:10,490 --> 00:30:13,490 Josh: all we got for this week there was a lot of craziness a 480 00:30:13,490 --> 00:30:16,610 Josh: lot of chaos all across the board all of our favorites open ai meta 481 00:30:16,610 --> 00:30:19,610 Josh: google apple they all got some screen time 482 00:30:19,610 --> 00:30:22,630 Josh: this week because they're all up to no good well 483 00:30:22,630 --> 00:30:25,730 Josh: i guess some some are up to some good it's a it's a mixed bag this week uh 484 00:30:25,730 --> 00:30:28,490 Josh: there was a good a circular good there was one last 485 00:30:28,490 --> 00:30:31,530 Josh: bit of news um on the topic of space and uh 486 00:30:31,530 --> 00:30:34,650 Josh: that is that jared isaacman has been nominated for head of nasa which 487 00:30:34,650 --> 00:30:38,270 Josh: is exciting because nasa i'm not sure if you realize this nasa at one point 488 00:30:38,270 --> 00:30:41,010 Josh: they sent people to outer space like they actually had rockets that went up 489 00:30:41,010 --> 00:30:46,230 Josh: into outer space and it functioned it was a functioning part of society uh which 490 00:30:46,230 --> 00:30:48,990 Josh: has unfortunately degraded over the years and the space program has kind of 491 00:30:48,990 --> 00:30:52,690 Josh: faltered to nothing which is i think where a lot of the enthusiasm around SpaceX come from. 492 00:30:53,270 --> 00:30:56,470 Josh: Thankfully, Jared is someone who is really passionate about space. 493 00:30:56,590 --> 00:30:58,230 Josh: In fact, he has been there twice. 494 00:30:58,550 --> 00:31:02,390 Josh: So he has a lot of experience. He understands how this works. 495 00:31:02,390 --> 00:31:07,730 Josh: And I'm hoping we'll provide a jolt into NASA to just make the space program more exciting again. 496 00:31:08,510 --> 00:31:12,890 Josh: They have the Artemis II program, I believe, which is supposed to be taking the 497 00:31:13,190 --> 00:31:18,350 Josh: a rocket to the moon fairly soon. So I'm just excited to see this new ambition happening with space. 498 00:31:18,830 --> 00:31:21,970 Josh: Congrats to Jared on a nice, a nice win. 499 00:31:22,410 --> 00:31:25,370 Josh: And that's pretty much it. That's everything for this week. If you enjoyed, 500 00:31:25,530 --> 00:31:28,790 Josh: as always, please do not forget to share with your friends, like, 501 00:31:29,130 --> 00:31:31,730 Josh: subscribe, drop a comment about what you want to hear about next. 502 00:31:32,330 --> 00:31:35,490 Josh: We are doing slightly better on Spotify. Thank you for subscribing. 503 00:31:35,650 --> 00:31:39,610 Josh: If you have not gone to Spotify and watch there and followed us there and consume 504 00:31:39,610 --> 00:31:40,970 Josh: content there please we do 505 00:31:40,970 --> 00:31:45,810 Ejaaz: Not have a government bailout so we need you we need you to bail us out when 506 00:31:45,810 --> 00:31:50,790 Ejaaz: you are the back stars you are the backstop turn on your notifications subscribe 507 00:31:50,790 --> 00:31:54,010 Ejaaz: to us wherever you are tell your friends to do it even if they don't listen 508 00:31:54,010 --> 00:31:55,770 Ejaaz: it would help us out so much. 509 00:31:55,770 --> 00:32:00,730 Josh: So with that i think we're done for the week um we're gonna go take a nice show 510 00:32:00,730 --> 00:32:04,630 Josh: weekend we'll be back bright and early next week for whatever hot news happens 511 00:32:04,630 --> 00:32:08,710 Josh: to come over the weekend so stay tuned and as always we will see you guys in 512 00:32:08,710 --> 00:32:10,650 Josh: the next one until next time peace 513 00:32:10,650 --> 00:32:11,150 Ejaaz: See you guys.