1 00:00:02,120 --> 00:00:07,360 Speaker 1: Zone Media. Okaybia, these fuckers are right for containment. This 2 00:00:07,400 --> 00:00:09,280 Speaker 1: is your better offline monologue for the week, and I'm 3 00:00:09,320 --> 00:00:22,880 Speaker 1: your host ed Ze truck. Now today, I'm going to 4 00:00:22,960 --> 00:00:24,759 Speaker 1: go over a question I get asked a lot. What 5 00:00:24,880 --> 00:00:28,120 Speaker 1: actually bursts a bubble? As I'm writing recording this, I'm 6 00:00:28,120 --> 00:00:30,760 Speaker 1: watching in video's stock wobble up and down. After analyst 7 00:00:30,840 --> 00:00:34,199 Speaker 1: TD Cohen revealed that Microsoft has canceled agreements for multiple 8 00:00:34,280 --> 00:00:37,640 Speaker 1: data centers. Cohen also mentioned that Microsoft had and I 9 00:00:37,720 --> 00:00:41,920 Speaker 1: quote Bloomberg pulled back on converting so called statements of qualifications, 10 00:00:42,080 --> 00:00:44,440 Speaker 1: which are the precursor to agreements that include things like 11 00:00:44,520 --> 00:00:47,839 Speaker 1: financing deals and payment structures, meaning that Microsoft does not 12 00:00:47,920 --> 00:00:50,479 Speaker 1: just canceled agreements, but made it clear they don't intend 13 00:00:50,560 --> 00:00:53,960 Speaker 1: to carry on building more. Microsoft responded to this by 14 00:00:54,000 --> 00:00:56,520 Speaker 1: saying that they were still sticking to their egregious, stupid 15 00:00:56,520 --> 00:00:58,920 Speaker 1: plan to spend more than eighty billion dollars on capital 16 00:00:58,960 --> 00:01:02,000 Speaker 1: expenditures in twenty twenty, with one note that they and 17 00:01:02,080 --> 00:01:06,240 Speaker 1: I quote may strategically pace or adjust their infrastructure in 18 00:01:06,280 --> 00:01:09,160 Speaker 1: some areas. It was one of those things that ostensibly 19 00:01:09,160 --> 00:01:11,120 Speaker 1: looks like an obvious statement of intent, but if you 20 00:01:11,160 --> 00:01:13,280 Speaker 1: read between the lines, you see plenty of wiggle room 21 00:01:13,400 --> 00:01:16,360 Speaker 1: for Microsoft to backtrack. This comes off the back of 22 00:01:16,440 --> 00:01:19,520 Speaker 1: last week's interview with Microsoft CEO sach And Adela with 23 00:01:19,600 --> 00:01:24,040 Speaker 1: Dwarkesh Dwarkesh podcast Pretty good, kind of bland stuff, where 24 00:01:24,040 --> 00:01:26,760 Speaker 1: he said that there will be overbuild of data centers 25 00:01:26,800 --> 00:01:29,080 Speaker 1: pursuing AI and this is the shit you want to 26 00:01:29,160 --> 00:01:31,560 Speaker 1: hear as an investor in the AI revolution. You want 27 00:01:31,560 --> 00:01:33,720 Speaker 1: to hear a guy being like, yeah, maybe people built 28 00:01:33,720 --> 00:01:36,920 Speaker 1: too much. Now, this caused both Microsoft and in videos 29 00:01:36,920 --> 00:01:38,880 Speaker 1: stocks to wobble. But as I sit here and write 30 00:01:38,920 --> 00:01:41,840 Speaker 1: this on Wednesday morning, the day of Nvidia's earnings, by 31 00:01:41,840 --> 00:01:43,880 Speaker 1: the way, both stocks appear to have recovered a tiny 32 00:01:43,920 --> 00:01:45,960 Speaker 1: bit and they're actually going off, though not by much, 33 00:01:46,000 --> 00:01:48,040 Speaker 1: and no doubt by the time I'm finished writing the 34 00:01:48,080 --> 00:01:50,680 Speaker 1: script things will have changed again, which is why I 35 00:01:50,800 --> 00:01:53,200 Speaker 1: tend not to do a lot of stock related stuff. 36 00:01:53,440 --> 00:01:56,320 Speaker 1: That and I'm not a financial analyst. Different kind of analyst, 37 00:01:56,360 --> 00:02:02,400 Speaker 1: I guess. But anyway, the magnificent seven stocks Apple, Microsoft, Amazon, Alphabet, Meta, Invidia, 38 00:02:02,440 --> 00:02:05,080 Speaker 1: and Tesla make up about thirty percent of the value 39 00:02:05,080 --> 00:02:06,920 Speaker 1: of the S and P five hundred, an index of 40 00:02:06,920 --> 00:02:09,480 Speaker 1: the five hundred largest companies in the American stock market, 41 00:02:09,800 --> 00:02:12,600 Speaker 1: with Nvidia being the golden goose, growing over nine hundred 42 00:02:12,680 --> 00:02:15,760 Speaker 1: percent since twenty twenty three thanks to being the one 43 00:02:15,800 --> 00:02:19,280 Speaker 1: company that sells the specialized GPUs needed to make generative 44 00:02:19,320 --> 00:02:22,200 Speaker 1: AI work, making up just under twenty percent of the 45 00:02:22,280 --> 00:02:25,480 Speaker 1: value of the Magnificent seven itself. As a result, the 46 00:02:25,520 --> 00:02:28,320 Speaker 1: markets are deeply dependent on the success of Invidio, which 47 00:02:28,360 --> 00:02:31,360 Speaker 1: in turn is dependent on the continued hyperscalar investment in 48 00:02:31,440 --> 00:02:35,240 Speaker 1: generative AI. One other worrying fact for you, in Vidia's 49 00:02:35,280 --> 00:02:38,120 Speaker 1: top customers like Microsoft, Google, and Meta make up more 50 00:02:38,120 --> 00:02:42,800 Speaker 1: than sixty percent of its revenues. Right now, the vibes 51 00:02:42,800 --> 00:02:46,079 Speaker 1: are rancid Tuesday and Wednesday's headlines, and this episode will 52 00:02:46,120 --> 00:02:49,280 Speaker 1: come out after in Vidia's earnings. So forgive me Echo 53 00:02:49,360 --> 00:02:53,040 Speaker 1: a deep market anxiety saying things like Invidia's earnings could 54 00:02:53,040 --> 00:02:55,600 Speaker 1: be bad, how to protect your portfolio, and why in 55 00:02:55,720 --> 00:03:00,000 Speaker 1: Video's earnings are so important to the entire stock market now. Admittedly, 56 00:03:00,160 --> 00:03:02,000 Speaker 1: a lot of these headlines were plucked from the popular 57 00:03:02,040 --> 00:03:04,560 Speaker 1: financial press, which, to put it mildly, can be kind 58 00:03:04,560 --> 00:03:07,560 Speaker 1: of dog shit. The content farms that churn out articles 59 00:03:07,600 --> 00:03:10,400 Speaker 1: where underneath the sensational clip bait title. You have a 60 00:03:10,400 --> 00:03:12,600 Speaker 1: thinly written body of texts that tries to persuade you 61 00:03:12,600 --> 00:03:15,000 Speaker 1: that a particular stock will pop or bomb. You can't 62 00:03:15,040 --> 00:03:17,000 Speaker 1: really read too much into them, but it's telling that 63 00:03:17,040 --> 00:03:20,119 Speaker 1: the same sentiments have appeared in more prestigious publications, where 64 00:03:20,120 --> 00:03:22,799 Speaker 1: standards are measurably higher than say, Seeking Alpha or The 65 00:03:22,840 --> 00:03:26,200 Speaker 1: Motley Fool. Even if earnings are good for Invidia, it's 66 00:03:26,200 --> 00:03:28,000 Speaker 1: hard to see how they'll be good enough to please 67 00:03:28,000 --> 00:03:30,399 Speaker 1: a market. They will have a tantrum if they can't 68 00:03:30,400 --> 00:03:33,560 Speaker 1: perpetually make the number go up. Invidia has beaten analyst 69 00:03:33,639 --> 00:03:37,120 Speaker 1: expectations eight straight quarters, and the continual expectation that they'll 70 00:03:37,120 --> 00:03:40,800 Speaker 1: do so is terrifying, especially when you consider that Invidia 71 00:03:40,840 --> 00:03:43,200 Speaker 1: has long held a reputation as a boom or bus 72 00:03:43,280 --> 00:03:46,720 Speaker 1: stock and osolits between peaks and valleys based on whatever 73 00:03:46,720 --> 00:03:50,240 Speaker 1: the latest tech trend is. There's no logical reason to 74 00:03:50,280 --> 00:03:53,080 Speaker 1: believe that Invidia can continue to grow at this ridiculous rate. 75 00:03:53,640 --> 00:03:56,560 Speaker 1: Right now, its market cap is over eleven percent of 76 00:03:56,720 --> 00:04:00,440 Speaker 1: America's gross domestic product. But because the market's like giant, 77 00:04:00,520 --> 00:04:03,800 Speaker 1: petulant trillion dollar babies, these are the demands put upon 78 00:04:03,800 --> 00:04:07,040 Speaker 1: in video and Jensen Wang. But even if in Vidia 79 00:04:07,120 --> 00:04:10,160 Speaker 1: shits the bed mightily tonight, it isn't obvious whether that 80 00:04:10,240 --> 00:04:12,440 Speaker 1: will burst the bubble, because so much of the stock's 81 00:04:12,520 --> 00:04:14,960 Speaker 1: value and the economy in general, if I'm honest, is 82 00:04:15,000 --> 00:04:17,560 Speaker 1: based on a mishmash of people pretending they understand numbers 83 00:04:17,600 --> 00:04:21,400 Speaker 1: and well vibes. In short, the bubble will not burst 84 00:04:21,400 --> 00:04:23,880 Speaker 1: in the sense that one big event will bring everything 85 00:04:23,920 --> 00:04:26,080 Speaker 1: to an abrupt end now, and Vidia is not going 86 00:04:26,080 --> 00:04:29,280 Speaker 1: to drop thirty percent in one day normal open MLA overnight. 87 00:04:29,360 --> 00:04:32,600 Speaker 1: That's just not how this works. What will happen if 88 00:04:32,640 --> 00:04:35,800 Speaker 1: the bubble bursts, which I genuinely believe it will, will 89 00:04:35,839 --> 00:04:38,200 Speaker 1: be a series of smaller calamities that chain together to 90 00:04:38,240 --> 00:04:40,359 Speaker 1: bring an end to this hype cycle. It'll be like 91 00:04:40,400 --> 00:04:43,360 Speaker 1: the Domino scene from viv and Vendetta, where Detective Finch 92 00:04:43,520 --> 00:04:46,160 Speaker 1: threads the needle between the disparate events that will eventually 93 00:04:46,200 --> 00:04:49,560 Speaker 1: lead to the ulster of Chancellor Sutler. I've referred in 94 00:04:49,600 --> 00:04:51,520 Speaker 1: the past in an episode called burst damage to the 95 00:04:51,600 --> 00:04:54,039 Speaker 1: pale horses of the AI Apocalypse, and one of them 96 00:04:54,080 --> 00:04:57,560 Speaker 1: was the reduction of capital expenditures by a major hyperscaler 97 00:04:57,680 --> 00:05:01,240 Speaker 1: because in doing so, the hyperscalers such as Microsoft would 98 00:05:01,279 --> 00:05:03,599 Speaker 1: be admitting that it's time to slow down investment in 99 00:05:03,680 --> 00:05:06,760 Speaker 1: revolutionary products that they allegedly claim in revolutionary and that 100 00:05:06,800 --> 00:05:10,840 Speaker 1: they're selling tons of right wrong. And by extension, by 101 00:05:10,839 --> 00:05:13,880 Speaker 1: the way, they're tacitly admitting that perhaps these revolutionary products 102 00:05:13,880 --> 00:05:17,520 Speaker 1: weren't actually that revolutionary. Kind of an obvious point. Now, 103 00:05:17,560 --> 00:05:20,320 Speaker 1: one argument against this being a pale horse is that 104 00:05:20,400 --> 00:05:22,760 Speaker 1: open ai has partnered with soft Bank and Oracle to 105 00:05:22,800 --> 00:05:25,599 Speaker 1: build out up to and that is what it's called, 106 00:05:25,880 --> 00:05:28,880 Speaker 1: up to five hundred billion dollars in data centers. And 107 00:05:28,920 --> 00:05:31,719 Speaker 1: the answer there is that this is right now pretty 108 00:05:31,760 --> 00:05:34,919 Speaker 1: much theoretical, and it's dependent on open ai raising another 109 00:05:35,040 --> 00:05:38,200 Speaker 1: nineteen billion dollars because they've committed that much to this project. 110 00:05:38,560 --> 00:05:41,040 Speaker 1: Another thing about this stargate thing, I get a lot 111 00:05:41,080 --> 00:05:44,440 Speaker 1: of emails about Donald Trump US government. They're not putting 112 00:05:44,480 --> 00:05:47,960 Speaker 1: any money into this. This is all private partnerships, and 113 00:05:48,120 --> 00:05:50,680 Speaker 1: soft Bank and open ai have given nineteen billion while 114 00:05:50,680 --> 00:05:52,840 Speaker 1: they're claiming they're going to give nineteen billion dollars each. 115 00:05:53,120 --> 00:05:55,760 Speaker 1: And open ai is now raising up to forty billion 116 00:05:55,800 --> 00:05:58,680 Speaker 1: dollars with twenty five billion or up to twenty five 117 00:05:58,680 --> 00:06:00,599 Speaker 1: billion dollars, of that coming from software, So it's like, 118 00:06:01,000 --> 00:06:04,719 Speaker 1: what actually is happening here? Anyway? If Microsoft was so 119 00:06:04,839 --> 00:06:07,880 Speaker 1: invested in the generative AI revolution, why would they reduce 120 00:06:07,920 --> 00:06:11,599 Speaker 1: their investments in it? Surely this supposedly incredible demand and 121 00:06:11,640 --> 00:06:14,039 Speaker 1: the incredible opportunities of generative AI would mean that they 122 00:06:14,160 --> 00:06:20,280 Speaker 1: needed more data centers, right right, So strange. Nevertheless, if 123 00:06:20,320 --> 00:06:22,599 Speaker 1: you see Google or Metal or Amazon pulled back on 124 00:06:22,640 --> 00:06:25,520 Speaker 1: capital expenditures, that is the bell ringing. But here are 125 00:06:25,560 --> 00:06:28,160 Speaker 1: some other pale horses to watch out for. If open 126 00:06:28,200 --> 00:06:30,599 Speaker 1: ai and Anthropic raise their prices, it means that they're 127 00:06:30,600 --> 00:06:35,400 Speaker 1: finally having to cover their ruinous, horrible, unprofitable and unsustainable expenses. 128 00:06:35,880 --> 00:06:38,960 Speaker 1: Open Ai spent nine billion dollars to lose five billion 129 00:06:39,000 --> 00:06:42,159 Speaker 1: dollars in twenty twenty four, and an Anthropic lost five 130 00:06:42,200 --> 00:06:45,520 Speaker 1: point six billion dollars and only made just over nine 131 00:06:45,560 --> 00:06:48,560 Speaker 1: hundred million dollars in the same period. All of their 132 00:06:48,560 --> 00:06:51,320 Speaker 1: products are deeply unprofitable, and thus any price increases are 133 00:06:51,320 --> 00:06:54,640 Speaker 1: assigned that they need money. They need money now, money 134 00:06:54,680 --> 00:06:59,360 Speaker 1: mean money. Now. Conversely, if you see any dramatic price decreases, 135 00:06:59,400 --> 00:07:02,200 Speaker 1: this isn't the necessarily a sign of improvements in efficiency. 136 00:07:02,440 --> 00:07:04,600 Speaker 1: Sooned Up, a shy of Google, just announced that Google's 137 00:07:04,640 --> 00:07:07,440 Speaker 1: Gemini codesyst is now free for up to one hundred 138 00:07:07,480 --> 00:07:10,480 Speaker 1: and eighty thousand code completions a month. This puts a 139 00:07:10,520 --> 00:07:14,000 Speaker 1: direct price pressure on both open Ai and Anthropic, who 140 00:07:14,160 --> 00:07:17,480 Speaker 1: just launched a new version of their Claude model, and Microsoft, 141 00:07:17,480 --> 00:07:20,840 Speaker 1: which owns GitHub and by extension, GitHub Copiler, which costs 142 00:07:20,840 --> 00:07:23,640 Speaker 1: twenty dollars per month and reportedly still loses twenty to 143 00:07:23,680 --> 00:07:27,080 Speaker 1: eighty dollars a month they use. It's insane, But if 144 00:07:27,120 --> 00:07:29,400 Speaker 1: prices are forced to drop, this is a bad sign. 145 00:07:29,520 --> 00:07:32,720 Speaker 1: These are very unprofitable businesses. Now, if you see anything 146 00:07:32,760 --> 00:07:36,880 Speaker 1: about any prominent AI companies Anthropic, Open Ai, Scale, Coheer, Perplexity, 147 00:07:36,960 --> 00:07:39,200 Speaker 1: so on, and so forth. If you see they're having 148 00:07:39,200 --> 00:07:42,160 Speaker 1: trouble raising money, that means venture capitalists are scared, which 149 00:07:42,200 --> 00:07:45,480 Speaker 1: means everybody should be scared. Similarly, if you hear any 150 00:07:45,480 --> 00:07:48,880 Speaker 1: discord within these companies, people living on mass layoffs, senior 151 00:07:48,880 --> 00:07:51,800 Speaker 1: executives fleeing, that is also a bad sign. It means 152 00:07:51,880 --> 00:07:54,240 Speaker 1: that they no longer have faith in these companies. And 153 00:07:54,320 --> 00:07:56,480 Speaker 1: now if they jump to another generator of AI company 154 00:07:56,520 --> 00:07:58,360 Speaker 1: that is also not a good sign. It's kind of 155 00:07:58,400 --> 00:08:01,320 Speaker 1: like jumping from the Titanic to the Ola Gay. And 156 00:08:01,520 --> 00:08:03,680 Speaker 1: we've already had a lot of this with open Ai, 157 00:08:03,760 --> 00:08:06,760 Speaker 1: with the department, the departures of chief researcher barretts Off 158 00:08:07,080 --> 00:08:10,200 Speaker 1: and Chief technology officer Mirror Murati. When that trickle becomes 159 00:08:10,240 --> 00:08:12,600 Speaker 1: a flood, you know that a calamity won't be far behind. 160 00:08:13,280 --> 00:08:15,640 Speaker 1: We've already had a few of these. But bullshit companies 161 00:08:15,680 --> 00:08:17,760 Speaker 1: raising bullshit money are one of the biggest tells that 162 00:08:17,800 --> 00:08:21,240 Speaker 1: we're in a bubble. Open ai co founder Iliosuitscaper is 163 00:08:21,280 --> 00:08:24,440 Speaker 1: currently raising a billion dollars at a fucking stupid valuation 164 00:08:24,480 --> 00:08:28,240 Speaker 1: of thirty billion dollars for his company's safe superintelligence. Now 165 00:08:28,280 --> 00:08:31,000 Speaker 1: you'd think with all that cash there'd be something to 166 00:08:31,040 --> 00:08:34,920 Speaker 1: show wrong. Oh, you imbecile, you dip shit, you more on, 167 00:08:35,000 --> 00:08:37,440 Speaker 1: you pig. He doesn't have a product at all, there's 168 00:08:37,440 --> 00:08:40,520 Speaker 1: no product, they're pre product, but they're worth thirty billion dollars. 169 00:08:40,800 --> 00:08:45,360 Speaker 1: What are we doing here? Similarly, former open AIICTO Mirror Maurati, 170 00:08:45,880 --> 00:08:48,880 Speaker 1: she has a new company called Thinking Machines, And you'd 171 00:08:48,920 --> 00:08:51,760 Speaker 1: think with the Wired headline saying they're finally ready to 172 00:08:51,800 --> 00:08:54,320 Speaker 1: announce what they're working on. They would announce I don't 173 00:08:54,360 --> 00:08:57,480 Speaker 1: know what they're working on, you'd be fucking wrong. Stephen 174 00:08:57,559 --> 00:09:00,640 Speaker 1: Levy should be absolutely bloody ashamed of himself. Steven's done 175 00:09:00,640 --> 00:09:02,640 Speaker 1: some good journalists me in the past, and he's also 176 00:09:02,679 --> 00:09:04,920 Speaker 1: done some projeneritive AI stuff that I wouldn't wipe my 177 00:09:04,960 --> 00:09:07,760 Speaker 1: ass whole with fucking disgusting, both the thing I just 178 00:09:07,800 --> 00:09:11,880 Speaker 1: said and the thing happening now. Anyway, if you're wondering, uh, 179 00:09:11,920 --> 00:09:15,400 Speaker 1: what this company does, they and I quote develop top 180 00:09:15,440 --> 00:09:18,280 Speaker 1: notch AI with an eye towards making it useful and accessible. 181 00:09:18,440 --> 00:09:20,720 Speaker 1: I wish I could get away with just saying nothing 182 00:09:20,760 --> 00:09:23,680 Speaker 1: and getting a big pile of money. Wait, is that 183 00:09:23,720 --> 00:09:26,720 Speaker 1: what a podcast is? Anyway? There are, of course other 184 00:09:26,760 --> 00:09:29,120 Speaker 1: pale horses, but the core problem with bubbles is that 185 00:09:29,120 --> 00:09:32,120 Speaker 1: they're based less on tangible events and more on everybody 186 00:09:32,240 --> 00:09:34,439 Speaker 1: sharing the same delusion, which in this case is that 187 00:09:34,559 --> 00:09:37,440 Speaker 1: large language models will turn into something magical somehow, and 188 00:09:37,559 --> 00:09:39,160 Speaker 1: the only way to make them do so is to 189 00:09:39,200 --> 00:09:42,439 Speaker 1: plow billions of dollars into a fucking furnace. One pale 190 00:09:42,720 --> 00:09:44,960 Speaker 1: horse would be a vibe shift where people stop talking 191 00:09:44,960 --> 00:09:48,320 Speaker 1: about open AI and generative AI and this breathless wank tone, 192 00:09:48,480 --> 00:09:51,160 Speaker 1: but rather with an air of disdain and skepticism that 193 00:09:51,720 --> 00:09:54,600 Speaker 1: in and of itself is pretty hard to measure. What 194 00:09:54,679 --> 00:09:57,720 Speaker 1: breaks narratives and pops bubbles is when the consensus shifts 195 00:09:57,720 --> 00:10:01,280 Speaker 1: from dreaming to cold, harsh reality. As investors realize there's 196 00:10:01,280 --> 00:10:03,520 Speaker 1: no growth to be had from generative AI, they'll turn 197 00:10:03,559 --> 00:10:06,040 Speaker 1: their backs on both the generative AI revolution and the 198 00:10:06,040 --> 00:10:09,000 Speaker 1: companies that pushed it, and if they ever turn on video, 199 00:10:09,080 --> 00:10:12,400 Speaker 1: that will accelerate the AI apocalypse. Because in Vidia, despite 200 00:10:12,400 --> 00:10:15,560 Speaker 1: the real overblown valuations, actually makes a profit and sells 201 00:10:15,559 --> 00:10:20,240 Speaker 1: real physical things, though their consumer graphics cards are currently 202 00:10:20,280 --> 00:10:23,959 Speaker 1: melting at the fifteen ninety not great but really though, 203 00:10:24,000 --> 00:10:26,480 Speaker 1: in Video is really the only company to make serious 204 00:10:26,480 --> 00:10:30,040 Speaker 1: money from generative AI, and if that narrative changes, if 205 00:10:30,040 --> 00:10:33,480 Speaker 1: people turn on video, it's not going to be great 206 00:10:34,200 --> 00:10:35,920 Speaker 1: either way. I'll be here to walk you through it.