WEBVTT - AI Is Worse Than The Dot Com Bubble: Part One 0:00:03.200 --> 0:00:03.560 Media. 0:00:04.920 --> 0:00:17.640 Hi, I'm d Zeitron and this is Better Offline. 0:00:19.840 --> 0:00:21.040 Welcome to dot Comweek. 0:00:21.040 --> 0:00:23.040 I hope you enjoyed my chat with Matt Rosof to 0:00:23.079 --> 0:00:25.040 kick it off, and today is the first of a 0:00:25.120 --> 0:00:27.840 four part special But why I think the AI bubble 0:00:27.880 --> 0:00:30.639 is that much worse than the dot com bubble. So 0:00:31.080 --> 0:00:33.600 for almost five years I put out a newsletter or 0:00:33.800 --> 0:00:37.159 article basically every week that addresses the underlying financial rat 0:00:37.440 --> 0:00:39.519 at the heart of the tech industry, whether it be 0:00:39.560 --> 0:00:43.280 the outright corruption the crypto industry, the bullshit of the metaverse, 0:00:43.560 --> 0:00:46.000 or the AI bubble that I won't shut up about. 0:00:46.479 --> 0:00:48.559 It's been some of the most challenging work I've ever done. 0:00:48.640 --> 0:00:51.320 I've been forced to learn accountancy, economics and the names 0:00:51.320 --> 0:00:54.600 of at least ten different guys i'd gladly see under 0:00:54.640 --> 0:00:59.480 a with a I'm not on accountant, forensic, or otherwise. 0:00:59.560 --> 0:01:02.000 I'm not a banker or a financial analyst. I've never 0:01:02.040 --> 0:01:04.480 worked as a venture capitalist or a hedge fund, or 0:01:04.520 --> 0:01:06.600 at a hedge fund. I guess that would be I'm 0:01:06.640 --> 0:01:09.160 just a guy with a laptop and a microphone. But 0:01:09.240 --> 0:01:11.360 I think I've picked up a great deal of knowledge 0:01:11.360 --> 0:01:14.040 in the last few years, and the clarity of starting 0:01:14.040 --> 0:01:16.200 from a position of oh God, what does that mean 0:01:16.520 --> 0:01:18.399 actually allowed me to see things in the way that 0:01:18.480 --> 0:01:22.120 many haven't, which is why I am so utterly horrified 0:01:22.160 --> 0:01:25.279 when I hear people flippantly say that the AI bubble 0:01:25.280 --> 0:01:27.760 will work out fine because it's just like the dot 0:01:27.800 --> 0:01:30.720 com bubble. While there are similarities, I need to be 0:01:30.800 --> 0:01:33.440 clear that I think the AI bubble is far, far worse, 0:01:33.440 --> 0:01:35.839 and the calamity that follow us would be far more destructive. 0:01:36.560 --> 0:01:38.560 Let's review the dot com. 0:01:38.360 --> 0:01:41.480 Bubble actually had two elements, the stable, sensible tech companies 0:01:41.520 --> 0:01:43.399 that actually made money, and then the likes of pets 0:01:43.440 --> 0:01:46.080 dot com and webvan, both ideas that would eventually find 0:01:46.080 --> 0:01:49.160 margin positive existence in the form of Dewey and Instacram. 0:01:49.440 --> 0:01:52.440 I'd also add the telecommunications companies onto the side of 0:01:52.480 --> 0:01:55.280 this as well, but technically they weren't making the websites. 0:01:55.320 --> 0:01:58.400 They were just building me into them. What buried these 0:01:58.440 --> 0:02:00.640 companies was their obsession with growth, with them a rush 0:02:00.680 --> 0:02:03.200 to take them public. The idea of ordering pet food 0:02:03.240 --> 0:02:05.440 or groceries online was one that would have made a 0:02:05.440 --> 0:02:07.480 lot more sense if more people were connected to the 0:02:07.520 --> 0:02:10.519 Internet or had faster connections in the year two thousand, 0:02:10.680 --> 0:02:13.680 only about fifty two percent of Americans were online, and well, 0:02:13.960 --> 0:02:15.919 both web van and pets dot Com were losing two 0:02:15.960 --> 0:02:17.560 dollars for every dollar of revenue. 0:02:17.760 --> 0:02:18.200 I don't know. 0:02:18.320 --> 0:02:20.440 That shit didn't make any sense, and it wouldn't have 0:02:20.520 --> 0:02:22.760 even if we had high speed internet. But I think 0:02:22.800 --> 0:02:24.840 by the middle of the two thousands we only had 0:02:24.840 --> 0:02:27.680 most people on at best four hundred kilobits per second, 0:02:28.040 --> 0:02:32.160 so we're talking websites that took time to load. Anyway, 0:02:32.280 --> 0:02:35.840 the economic sourceer didn't make sense. Now the AI bubble 0:02:35.840 --> 0:02:38.480 differs in so far as that what we would have 0:02:38.560 --> 0:02:42.239 once considered stable sensible tech companies are acting irrationally, having 0:02:42.280 --> 0:02:44.840 doubled or in the case of Microsoft, tripled the amount 0:02:44.880 --> 0:02:47.359 of hardware known as PP and E that they operate 0:02:47.400 --> 0:02:49.440 in just a couple of years, with no signs of 0:02:49.480 --> 0:02:53.120 slowing down. These companies are racking up entering into multi 0:02:53.160 --> 0:02:56.600 decade long lease agreements, and it accumulated so much hardware 0:02:56.600 --> 0:02:59.760 that the depreciation will erode any profitability for the short 0:02:59.840 --> 0:03:03.560 term medium term future. For example, Microsoft went from having 0:03:03.639 --> 0:03:06.920 around eighty eight billion dollars in PP and E that's property, 0:03:06.919 --> 0:03:09.280 plants of equipment which is where the GPUs are, by 0:03:09.320 --> 0:03:11.560 the way, in the beginning of twenty twenty three, to 0:03:11.639 --> 0:03:14.960 a remarkable two hundred and thirty billion dollars in pp 0:03:15.080 --> 0:03:18.680 and E as of its last quarter. And depreciation, by 0:03:18.720 --> 0:03:21.480 the way, is not it's not a cash expense. They've 0:03:21.480 --> 0:03:23.800 already bought the GPUs and the service and the like. 0:03:24.120 --> 0:03:26.799 What it does, however, is sit there because they spread 0:03:26.840 --> 0:03:29.320 it out over five to six years. I think it's 0:03:29.360 --> 0:03:31.760 five and a half from Microsoft. They spread it out, 0:03:31.800 --> 0:03:35.400 so it eats into that net income side. So instead 0:03:35.400 --> 0:03:38.120 of they'd taken the immediate upfront hit, they spread it out. 0:03:38.360 --> 0:03:41.840 So this means that each quarter from here until fucking eternity, 0:03:41.880 --> 0:03:44.560 at this point, Microsoft is going to be taking billions 0:03:44.600 --> 0:03:48.520 of dollars of appreciation charges. Now, while the dot com 0:03:48.560 --> 0:03:51.280 bubble was wasteful, it didn't involve the largest and most 0:03:51.320 --> 0:03:53.760 well respected tech firms in the world accruing hundreds of 0:03:53.760 --> 0:03:56.560 billions of dollars of GPUs that are absolutely a year 0:03:56.680 --> 0:04:00.400 or so after being installed. It also didn't involve billions 0:04:00.440 --> 0:04:03.880 in operational expenses or acres of data centers. In fact, 0:04:03.960 --> 0:04:08.680 there's really no comparison. Fiber doesn't compare at all the 0:04:08.720 --> 0:04:10.800 scale of the fiber will get to it. But I 0:04:10.840 --> 0:04:13.960 just want to be clear that fiber is also something 0:04:14.000 --> 0:04:16.880 that you can use on many things. Fiber is something 0:04:16.920 --> 0:04:20.000 that another company could come up and run with. Aigpus 0:04:20.040 --> 0:04:23.320 are extremely limited in their outcomes. I've been over this before. 0:04:23.680 --> 0:04:26.000 Not going to repeat myself. Perhaps I'll do it more 0:04:26.040 --> 0:04:29.720 in depth episode on that in the future. But look, 0:04:29.839 --> 0:04:31.680 in many ways, I fear the AI bubble is going 0:04:31.720 --> 0:04:33.880 to be worse than the dot com bubble. It's going 0:04:33.920 --> 0:04:36.400 to make it look small, and we are already seeing 0:04:36.400 --> 0:04:39.600 some worrying signs. Now, let's start somewhere simple though. 0:04:39.880 --> 0:04:40.400 Layoffs. 0:04:41.000 --> 0:04:44.159 Fun fact, even during the dot com bubble, Microsoft barely 0:04:44.200 --> 0:04:46.599 had any layoffs, with the company growing its headcount even 0:04:46.600 --> 0:04:49.160 as the tech industry around it was consumed in flames. 0:04:50.480 --> 0:04:52.720 The biggest round of firings I could find during the 0:04:52.800 --> 0:04:55.599 nineteen ninety eight to two thousand and eight period was 0:04:55.600 --> 0:04:58.160 between August twenty twenty four and January twenty twenty five, 0:04:58.320 --> 0:05:00.440 when it can seventy six members of its ex Box 0:05:00.480 --> 0:05:02.560 group as well as one hundred and fifty seven test 0:05:02.600 --> 0:05:05.359 engineers whose jobs were offshore to India. That gives you 0:05:05.400 --> 0:05:07.320 a grand total of two hundred and thirty three people. 0:05:09.120 --> 0:05:11.800 For comparison, Microsoft laid off six thousand people in May 0:05:11.839 --> 0:05:14.880 of last year, or three percent of its global workforce. 0:05:15.560 --> 0:05:18.400 Now that's not to say that people at other companies 0:05:18.400 --> 0:05:21.760 were equally fortunate. In two thousand and one, some Microsystems 0:05:21.800 --> 0:05:24.839 called ninety percent of its workforce or thirty nine hundred jobs. 0:05:25.040 --> 0:05:27.359 That was the first of many at that particular company. 0:05:27.720 --> 0:05:31.080 In two thousand and one, the telecommunications sector cut and 0:05:31.120 --> 0:05:34.320 I am not shitting you here three hundred and seventeen thousand, 0:05:34.360 --> 0:05:37.120 seven hundred and seventy seven rolls, with the computer industry 0:05:37.200 --> 0:05:39.719 second rank for that year, shedding one hundred and fifty 0:05:39.720 --> 0:05:44.240 three thousand and nine hundred and fifty two positions. Similarly, 0:05:44.400 --> 0:05:47.080 Amazon saw it stocked tumble in January two thousand when 0:05:47.120 --> 0:05:48.880 it was discovered that it laid off one hundred and 0:05:48.960 --> 0:05:51.840 fifty people, two percent of its workforce at the time, 0:05:52.000 --> 0:05:54.360 and would lay off another thirteen hundred a year later, 0:05:54.520 --> 0:05:58.320 or fifteen percent of its workforce. Now, that, by the way, 0:05:58.440 --> 0:06:00.960 sounds like a lot. And also I think it's adorable. 0:06:01.000 --> 0:06:03.960 Back when the stock market cared, when companies did layoffs, 0:06:04.720 --> 0:06:07.080 they don't today. And Amazon indeed is a very different 0:06:07.120 --> 0:06:09.839 company today, going from a fledgling digital book seller to 0:06:09.839 --> 0:06:12.279 one of the largest retailers and cloud storage providers in 0:06:12.279 --> 0:06:15.400 the world, and a very profitable one. And that Yet 0:06:15.600 --> 0:06:19.800 this week Amazon laid off another fourteen thousand people, ten 0:06:19.839 --> 0:06:23.359 percent of its corporate workforce, around three months after laying 0:06:23.360 --> 0:06:26.360 off fourteen thousand more people in October twenty twenty five. 0:06:27.240 --> 0:06:29.080 Why well bids two things. 0:06:29.600 --> 0:06:32.039 First, it's the scourge of Jack welch, go and listen 0:06:32.080 --> 0:06:34.359 to the shareholder supremacy from twenty twenty four if you 0:06:34.440 --> 0:06:36.800 like that one, and the idea that laying people off 0:06:37.120 --> 0:06:40.280 well at boosts profits, which Wall Street loves now. Secondly, 0:06:40.320 --> 0:06:43.000 it's because Amazon is buying billions of dollars of GPUs, 0:06:43.080 --> 0:06:46.680 both in Vidia and directly from TSMC for its trainium 0:06:46.680 --> 0:06:52.200 custom chips. That is TSMC Taiwan Semiconductor Manufacturing Corporation they 0:06:52.360 --> 0:06:56.400 build basically every chip, and Amazon they build them internally 0:06:56.480 --> 0:07:00.000 using a company called Annapurna Labs. It's a whole thing 0:07:00.000 --> 0:07:01.840 saying it doesn't really mean much. They're not as good 0:07:01.839 --> 0:07:04.640 as vidious GPUs, but nevertheless they keep fucking that chicken. 0:07:05.120 --> 0:07:07.440 And this is a massive burden on its earnings. These 0:07:07.520 --> 0:07:10.360 chips have their costs spread out, like I mentioned as 0:07:10.400 --> 0:07:13.360 a depreciation charge, dragging down the profits that Amazon can 0:07:13.400 --> 0:07:16.160 report on its earnings in the process. Well, this isn't 0:07:16.160 --> 0:07:18.600 the problem when you buy one or two or ten GPUs. 0:07:18.760 --> 0:07:20.720 It becomes one when you have hundreds of thousands of 0:07:20.760 --> 0:07:38.440 the fuckers. Amazon had sixteen point seven billion dollars of 0:07:38.520 --> 0:07:42.520 depreciation charges and its last earnings. That's a great deal. 0:07:43.440 --> 0:07:45.720 I think their net revenue is like twenty two to 0:07:45.760 --> 0:07:50.480 twenty four billion dollars, not great. In similar terms, today's 0:07:50.520 --> 0:07:52.920 layoffs are happening not because the companies are in trouble, 0:07:53.040 --> 0:07:54.720 but because they want to boost their profits at a 0:07:54.760 --> 0:07:57.360 time when AI services aren't providing a profit and I 0:07:57.400 --> 0:08:00.480 don't think they ever will. Nevertheless, these layoffs are a sign 0:08:00.520 --> 0:08:03.600 of something that these companies know that revenue growth isn't 0:08:03.680 --> 0:08:06.680 keeping pace with their ruinous expenditures. The only jobs that 0:08:06.720 --> 0:08:09.200 AI is taking are those that the hyperscaler cuts to 0:08:09.280 --> 0:08:10.720 keep paying for its existence. 0:08:11.640 --> 0:08:13.680 In any case, people really need. 0:08:13.480 --> 0:08:16.200 To recognize that the dot com bubble wasn't some small 0:08:16.240 --> 0:08:19.040 event that could be casually waved off, and other things 0:08:19.080 --> 0:08:21.560 that happen tell the story of what might before anyone 0:08:21.640 --> 0:08:24.520 left in or invested in the tech industry when it does. 0:08:25.640 --> 0:08:28.360 In two thousand and one, VC funds raised sixty three 0:08:28.400 --> 0:08:30.800 percent less than the previous year, and the amount of 0:08:30.800 --> 0:08:33.800 money going to startups, not just technology companies, dropped by 0:08:33.800 --> 0:08:37.400 slightly more. Then there's the stock marketing. If you held 0:08:37.440 --> 0:08:39.600 stocks in the company like pets dot Com were at 0:08:39.600 --> 0:08:43.200 its highest relatively modest four hundred million dollars, or Webvam 0:08:43.200 --> 0:08:45.760 with a peak market cap of eight billion dollars, you 0:08:45.800 --> 0:08:49.200 were completely fucking screwed. Those companies went to zero or 0:08:49.240 --> 0:08:52.440 near enough. Even those who invested in established firms like 0:08:52.480 --> 0:08:55.840 Microsoft and Sun faced their own haircuts, with Microsoft's share 0:08:55.840 --> 0:08:57.760 price dropping from a high of one hundred and twenty 0:08:57.800 --> 0:09:00.679 dollars to as little as forty dollars. Adjusted for the 0:09:00.720 --> 0:09:03.079 two stock splits that happened in nineteen ninety nine and 0:09:03.120 --> 0:09:07.079 two thousand and three, it would take around seventeen years 0:09:07.400 --> 0:09:10.400 for Microsoft to reach its original dot com bubble highs. 0:09:11.080 --> 0:09:12.920 The fact is, the dot com bubble fucked a lot 0:09:12.960 --> 0:09:16.319 of people. The tech sector's share of employment wouldn't surpass 0:09:16.360 --> 0:09:20.480 the levels of the bubble era until twenty fifteen. In practice, 0:09:20.520 --> 0:09:23.720 this meant that newly graduated software engineers couldn't find much work, 0:09:23.880 --> 0:09:26.120 and what work they could find offered stagnant wages that 0:09:26.160 --> 0:09:29.840 didn't keep pace with inflation. That's the thing about the 0:09:29.840 --> 0:09:31.840 dot com bubble. We like to think if it as 0:09:31.880 --> 0:09:33.920 something that happened around the turn of the millennium, but 0:09:33.960 --> 0:09:36.080 in practice it took until two thousand and four for 0:09:36.120 --> 0:09:38.960 the contraction and the tech jobs market to finally bottom out. 0:09:39.800 --> 0:09:43.599 Existing workers face threats not simply from economic related cutbacks 0:09:43.720 --> 0:09:45.880 making the company do the same with the resources they 0:09:45.880 --> 0:09:49.120 already have, but also from outsourcing, which grew exponentially in 0:09:49.160 --> 0:09:54.160 popularity in the early two thousands. This incidentally, was around 0:09:54.200 --> 0:09:57.240 the time that China entered the World Trade Organization, meaning 0:09:57.280 --> 0:09:59.360 that in addition to cuts in the service related the 0:09:59.600 --> 0:10:02.679 sector the economy, a lot of high tech manufacturing jobs 0:10:02.679 --> 0:10:05.640 went out the door too. The point I'm making is 0:10:05.679 --> 0:10:07.440 that the dot com bubble was bad, and that the 0:10:07.440 --> 0:10:09.720 second order effects, the things that didn't get as many 0:10:09.720 --> 0:10:13.560 headlines or as much pop culture weight, were really truly gruesome. 0:10:14.679 --> 0:10:16.319 Now there's a great line from The Big Shot. I 0:10:16.360 --> 0:10:19.240 always find myself returning to. It's when Ben rick At, 0:10:19.240 --> 0:10:21.680 the retired finance guy who came back in from the 0:10:21.679 --> 0:10:24.880 cold in order to help Brownfield Capital short the mortgage market, 0:10:25.040 --> 0:10:27.599 castigates the two young founders for celebrating what would be 0:10:27.640 --> 0:10:28.720 the trade of the century. 0:10:29.080 --> 0:10:32.720 If we're right, people lose homes, people lose jobs, people 0:10:32.760 --> 0:10:34.920 whose retirement savings, people lose pensions. 0:10:35.720 --> 0:10:37.600 You know what I hate about fucking banking. It reduces 0:10:37.600 --> 0:10:38.760 people to numbers. Here's a number. 0:10:38.760 --> 0:10:41.160 Every one percent, unemployment goes up, forty thousand people died. 0:10:41.200 --> 0:10:43.800 Did you know that It's so fucking easy to talk 0:10:43.800 --> 0:10:46.760 about this period with statistical data, to talk about layoffs 0:10:46.760 --> 0:10:49.720 and abstract terms like tens of thousands or hundreds of 0:10:49.720 --> 0:10:52.920 thousands of jobs, or to say how the stock market contracted, 0:10:52.920 --> 0:10:55.920 but things would work out okay, that everybody would be 0:10:56.000 --> 0:10:59.000 fine in the future, and thus the fuck ups today 0:10:59.000 --> 0:11:02.160 would be okay. I need to be clear that anybody 0:11:02.160 --> 0:11:04.520 who traded into the dot com bubble got washed out, 0:11:04.679 --> 0:11:07.880 and similarly, anybody who believed story after story about the 0:11:07.920 --> 0:11:11.920 eternal growth and telecommunications or web startups lost almost everything, 0:11:12.080 --> 0:11:15.360 if not everything they put in. Anybody who worked for Enron, 0:11:15.640 --> 0:11:18.640 the people that didn't know about the illegal bullshit happening, 0:11:18.840 --> 0:11:22.480 lost every cent of their stock based pension. Unemployment swelled 0:11:22.480 --> 0:11:24.320 to a peak of six point three percent in June 0:11:24.360 --> 0:11:26.480 two thousand and three, and the NASDAK lost nearly eight 0:11:26.480 --> 0:11:28.520 percent of its value from its peak in the year 0:11:28.559 --> 0:11:32.080 two thousand. Anyone saying that this is just like the 0:11:32.120 --> 0:11:34.559 dot com bubble as some sort of defense of the reckless, 0:11:34.880 --> 0:11:37.920 monstrous expenditures of the AI bubble is trying to find 0:11:37.960 --> 0:11:42.480 rationalizations for irrational, reckless actions. These are likely privileged people 0:11:42.480 --> 0:11:44.720 who define at the time were of found ways to 0:11:44.760 --> 0:11:47.439 look back at a time of incredible suffering and believe 0:11:47.520 --> 0:11:49.280 that because things are better today, that. 0:11:49.360 --> 0:11:51.480 All of it was worthwhile. It wasn't. 0:11:52.320 --> 0:11:54.520 I know that the fiber build out led to something, 0:11:54.559 --> 0:11:56.840 and I'll get to that in the future episode, I promise. 0:11:57.400 --> 0:12:00.600 But just because that happened doesn't make any this worthing. 0:12:01.360 --> 0:12:03.480 All of the wasted money in that time could have 0:12:03.480 --> 0:12:06.360 been spread out over a more thoughtful period. It wasn't, 0:12:06.400 --> 0:12:09.559 and people suffered as a result. I'm going to talk 0:12:09.600 --> 0:12:12.160 about this more next episode. Thank you for listening. 0:12:20.679 --> 0:12:22.359 Thank you for listening to Better Offline. 0:12:22.480 --> 0:12:24.920 The editor and composer of the Better Offline theme song 0:12:25.000 --> 0:12:27.640 is Matasowski. You can check out more of his music 0:12:27.640 --> 0:12:31.320 and audio projects at Mattasowski dot com, m A T 0:12:31.320 --> 0:12:31.920 T O. 0:12:32.120 --> 0:12:35.640 S O w Ski dot com. 0:12:35.679 --> 0:12:38.000 You can email me at easy at better offline dot 0:12:38.040 --> 0:12:40.240 com or visit better offline dot com to find more 0:12:40.280 --> 0:12:43.680 podcast links and of course, my newsletter. I also really 0:12:43.679 --> 0:12:46.000 recommend you go to chat dot Where's Youreed dot at 0:12:46.000 --> 0:12:48.680 to visit the discord, and go to our slash Better. 0:12:48.480 --> 0:12:51.640 Offline to check out our reddit. Thank you so much 0:12:51.679 --> 0:12:55.520 for listening. Better Offline is a production of cool Zone Media. 0:12:55.640 --> 0:12:58.480 For more from cool Zone Media, visit our website cool 0:12:58.559 --> 0:12:59.880 Zonemedia dot com. 0:13:00.200 --> 0:13:03.000 Check us out on the iHeartRadio app, Apple Podcasts, or 0:13:03.000 --> 0:13:04.400 wherever you get your podcasts.