Ed Zitron (Better Offline Podcast Host) (1:33)
Call Zone Media Hi, I'm Ed Zitron and this is better offline. Welcome to Dot Com Week. I hope you enjoyed my chat with Matt Rozoff to kick it off and today is the first of a four part special. But why? I think the AI bubble is that much worse than the dot com bubble. So for almost five years I put out a newsletter or article basically every week that addresses the underlying financial rot at the heart of the tech industry. Whether it be the outright corruption of the crypto industry, the bullshit of the metaverse, or the AI bubble that I won't shut up about. It's been some of the most challenging work I've ever done. I've been forced to learn accountancy, economics and the names of at least 10 different guys I gladly see under a with a. I'm not an accountant, forensic or otherwise. I'm not a banker or a financial analyst. I've never worked as a venture capitalist or a hedge fund. Or at a hedge fund. I guess that would be. I'm just the guy with a laptop and a microphone. But I think I've picked up a great deal of knowledge in the last few years and the clarity of starting from a position of oh God, what does that mean? Actually allowed me to see things in the way that many haven't. Which is why I'm so utterly horrified when I hear people flippantly say that the AI bubble will work out fine because it's just like the dot com bubble. While there are similarities, I need to be clear that I think the AI bubble is far far worse and the calamity that follows will be far more destructive. Let's review. The dot com bubble actually had two elements. The stable sensible tech companies that actually made money. And then the likes of Pets.com and Webvan, both ideas that would eventually find margin positive existence in the form of Chewy and Instacart. I'd also add the telecommunications companies onto the side of this as well. But technically they weren't making the websites, they were just building the Internet. What buried these companies was their obsession with growth and a rush to take them public. The idea of ordering pet food or groceries online was one that would have made a lot more sense if more people were connected to the Internet or had faster connections. In the year 2000, only about 52% of Americans were online and well, both Webvan and Pets.com were losing $2 for every dollar of revenue. I don't know. That shit didn't make any sense and it wouldn't have even if we had high speed Internet. But I think by the middle of the 2000s we only had most people on at best 400kbps. So we're talking websites that took time to load anyway. The economics also didn't make sense. Now the AI bubble differs insofar as that what we would have once considered stable, sensible tech companies are acting irrationally. Having doubled, or in the case of Microsoft, tripled the amount of hardware known as PP&E that they operate in just a couple of years with no signs of slowing down. These companies are racking up debt, entering into multi decade long lease agreements and accumulated so much hardware that the depreciation will erode any profitability for the short term medium term future. For example, Microsoft went from having around $88 billion in PP and E, that's property, plants and equipment, which is where the GPUs are by the way, in the beginning of 2023 to a remarkable $230 billion in PP&E as of its last quarter. And depreciation, by the way, is not. It's not a cash expense. They've already bought the GPUs and the servers and the like. What it does, however, is sit there because they spread it out over five to six years. I think it's five and a half for Microsoft. They spread it out so it eats into that net income side. So instead of taking the immediate upfront hit, they spread it out. So this means that each quarter from here until fucking eternity at this point, Microsoft is going to be taking billions of dollars of depreciation charges. Now while the dot com bubble was wasteful, it didn't involve the largest and most well respected tech firms in the world, accruing hundreds of billions of dollars of GPUs that were obsolete a year or so after being installed. It also didn't involve billions in operational expenses or acres of data centers. In fact, there's really no comparison. Fiber doesn't compare at all the scale of the fiber, but we'll get to it. But I just want to be clear that fiber is also something that you can use on many things. Fiber is something that another company could come up and run with. AI GPUs are extremely limited in their outcomes. I've been over this before, not going to repeat myself. Perhaps I'll do a more in depth episode on that in the future. But look, in many ways I fear the AI bubble is going to be worse than the dot com bubble. It's going to make it look small. And we are already seeing some worrying signs. Now let's start somewhere simple though. Layoffs. Fun fact. Even during the dot com bubble, Microsoft barely had any layoffs. With the company growing its headcount even as the tech industry around it was consumed in flames. The biggest round of firings I could find during the 1998-2008 period was between August 2024 and January 2025, when it can 76 members of its Xbox group as well as 157 test engineers whose jobs were offshored to India. That gives you a grand total of 233 people. For comparison, Microsoft laid off 6,000 people in May of last year, or 3% of its global workforce. Now that's not to say that people at other companies were equally fortunate. In 2001, some microsystems culled 90% of its workforce or 3,900 jobs. That was the many at that particular company. In 2001 the telecommunications sector cut, and I'm not shitting you here, 317,777 roles with the computer industry second ranked for that year, shedding 153,952 positions. Similarly, Amazon saw its stock tumble in January 2000 when it was discovered that it laid off 150 people, 2% of its workforce at the time, and would lay off another 1300 a year later, or 15% of its workforce. Now that, by the way, sounds like a lot. And also, I think it's adorable, back when the stock market cared when companies did layoffs, they don't today. And Amazon indeed is a very different company today, going from a fledgling digital bookseller to one of the largest retailers and cloud storage providers in the world, and a very profitable one at that. Yet this week, Amazon laid off another 14,000 people, 10% of its corporate workforce, around three months after laying off 14,000 more people in October 2025. Why? Well, it's two things. First, it's the scourge of Jack Welch. Go and listen to the shareholder supremacy from 2024, if you like that one. And the idea that laying people off, well, it boosts profits, which Wall street loves. Now. Secondly, it's because Amazon is buying billions of dollars of GPUs, both from Nvidia and directly from TSMC for its Trainium custom chips. That is tsmc, Taiwan Semiconductor Manufacturing Corporation. They build basically every chip and Amazon, or they build them internally using a company called Annapurna Labs. It's a whole thing. It doesn't really mean much. They're not as good as Nvidia's GPUs, but nevertheless, they keep fucking that chicken. And this is a massive burden on its earnings. These chips have their costs spread out, like I mentioned, as a depreciation charge, dragging down the profits that Amazon can report on its earnings in the process. While this isn't a problem when you buy one or two or 10 GPUs, it becomes one when you have hundreds of thousands of the fuckers.
