[0:00]I think that the majority of these projects are flimflam. I think they're vaporware. You have to wonder in a year whether you don't just see a crash in the pricing of Blackwell GPUs, if indeed all of them even get installed. The current conversion of hyperscalers, the big tech companies like Meta, Amazon, Microsoft, and Google. They've gone from these asset light cash machines to asset heavy behemoths rolling around in their own filth. Ever building construction. I'm not sure where the real economy is here. I don't, you've got GPUs being sold for data centers that aren't getting built. You've got data centers being built for customers that can't pay for them. You've got an industry claiming it's the biggest, most hugest thing ever. But when you look at the actual user numbers, pathetic.
[0:54]On the tech report with me today is writer of where's your Ed at and the host of the Batchelor Offline podcast. Ed Zitron, thanks for coming on again. What's up? How you doing? I'm very good, thank you. I read your newsletter on Tuesday and it painted a pretty damning picture of their AI buildup. Between data centers, construction figures kind of on one side and then the the power that kind of just isn't there for the most part, anyway. Let's start off with the fact that only a third of the US data center pipeline is currently being built. Do you think that the buildout has hit a physical ceiling in terms of what they're able to make in one go? Or is there simply not enough consumer demand for AI at the moment to motivate faster building? Little A, little B. So, here are the numbers. My estimate is there is only about 3 gigawatts of critical IT load, so GPUs and the associated bits and bobs installed in the US in uh 2025. Siteline Climate, however, found that there's a somewhere between 190 and 240 gigawatts of capacity being built. Being built, as in announced, but only 5 gigawatts are actually under construction worldwide. So, perhaps this is the, I don't know if this is demand related, but I think it's a reality related in the sense that construction is hard. The concept of a hyperscale data center, something over 50 megawatts, is only a few years old, like we have had, you don't really have actually 100 megawatt data centers. You have campuses of smaller ones, but even then, very new concept. The reality is that construction is hard, and building for data centers for AI is very difficult. Because of the cooling requirements, the bigger they are, the more heat that needs to be got rid of. It's very difficult to do, but also, I think there's some dodgy stuff happening. I think that there's dodgy building happening, because you're talking like 200 odd gigawatts. There's only 5 gigawatts that's actually being constructed. Siteline said that 16 gigawatts is meant to come on this year. I think that the majority of these projects are flimflam. I think they're vaporware and I actually am a premium this week, I went into it. I looked up some of the biggest projects in in the entirety of America, and almost all of them that I found were either not under construction, had only just broken ground or literally didn't exist. Ferramiy America, 11 gigawatt buildout in Amarillo, Texas, for example. Crazy stat. Uh yeah, 11 gigawatts, they were meant to be getting up and running sometime in 2026. Except as of February 2026, the whole thing is stopped because they didn't have the funding or the permitting. That is a publicly traded stock. Ferramiy America, uh Stargate, Abilene, two buildings out of eight are built. Was meant to be done at the end of 2026, then the middle of 2026, sorry, the end of 2025, middle of 2026, end of 2026. It's not going to be done before 2027. In fact, I don't even think it gets done at all. I think the whole thing's a Crockett nothing. Port Washington, Wisconsin, they have managed to put up, and I'm not kidding as of a month ago, a single steel beam. Just one. The picture's great. The head of infrastructure Open AI posted it. It's just three blocks standing around a single steel beam with a bloke walking up to it. I grew up in West London. I've seen a lot of these projects. A lot of blokes standing around with their hands on their hips, goes, oh, mate. Or, I don't know how you're going to get a GPU in there, mate. It's going to it's going to need it's going to need another month to build that, mate. Yeah, it's, I think what it is is that you've got a lot of construction money sloshing around and going nowhere. In America, only about 28 billion dollars was spent on data center construction last year. For what is meant to be hundreds of gigawatts worth. Power is definitely a bottleneck, but the actual real bottleneck is just reality. These things take a long while to build. I don't think they're fully capitalized. And I think we're going to see in the next year or two, just announcement after announcement of these things getting canned. Can you just explain, you go into absorption in in your newsletter. Can you just explain a little bit about what that is in this context of the particularly the flimflam that you're talking about? So absorption is data centers that come online and produce revenue. And CBRE, they, they said about two and a half gigawatts worth got absorbed in the in 2025 in America in the eight largest markets, which account for most data center capacity. I went around looking for the rest of the markets, you've got growing markets like Southern Virginia, Columbus, Ohio. And what's funny is, when I looked it up, lots of people are like, well, if you don't include those, it's not the full amount. Except, I could find very little actual stuff that got built there. Tons of announced stuff. There's, if you count all of the announced data centers, you could cover the entire country in future laser tag arenas. It's ridiculous. Absorption is the only number that matters, though, because when you hear under construction, under construction can mean literally everything from a single steel beam to a near complete building. And these companies, they like to mislead people, you saw it with Stargate Abeline in particular, they'll say, it's come online. When they say it's come online, click through to the article and make sure they're not just talking about one bit of it coming online, because that's always the case. So absorption is the only real measure of AI buildout. You hear an announcement of a data center, think two or three or four years in the future. Because I can't find a single data center project announced last year or even in 2024 that appears to have been finished. Even even if everything was on time, what about the power? Because you mentioned it a second ago. That about 60% of the data centers that are being built, don't even really have a power source for them ready. So that's not just they've basically just connected the wires up, but there's nothing to even turn them on. And it's a bit worse than that as well, because they and all of these articles about power of such weasel wording. Long story short, 60% of the data centers just don't have an answer for where the power's coming from. That does not mean that the 40% actually has power. It just means that they've signed contracts with power suppliers theoretically. 60% of them like, yeah, we're going to build it, we'll work out the power later. Yehaw. We don't need to bother, we don't need to worry about these things. You've got a lot of people claiming they're also going to do behind the meter, which is just their own power. And uh let me tell you, if I was building something right now, and I was using liquefied natural gas, I'd be a little bit concerned that, I think it, Saudi Arabia called force majeure, just saying that they can't supply people. It will take three to five years to restore the availability of natural gas to where it was before the war in Iran. So, yeah, just good luck to those guys. But nevertheless, that 60% is an insane figure. It just, it's just like a bunch of people that are like, yeah, we'll work it out in the future. How? Where? Like, what do you mean? Power is so, like these data centers are not getting built even before the power issues. You get to this point where when you, when you really think about it, none of this is real. Like, they're talking about they can't build the data centers, they don't have the power. When asked about the power, they're just like, nah, yeah, you know, we'll work it out, mate. It'll be sometime in the future, but they're not working it out. The power is not getting built in to even like 10% of what's necessary, and even if it does come online, that doesn't mean the power's getting to the data center. did some maths recently that for for a gigawatt of of power, if you were to just build a solar farm next to your data center, it would be about somewhere between 7 and 10 square miles, which is massive. You'd need, I think, Manhattan to produce about 2 gigawatts. So if we think about the time it's going to take to build all of these things and and the the power question, which is just there. What's happening to all of the insane amounts of GPUs, which are just kind of waiting, I presume? I don't know. I'm going to be completely honest. I worked it out that it takes about six months to install a single quarter of NVIDIA GPUs. So about two years for every year's worth of sales at this current rate. NVIDIA is saying they're going to sell even more of these bloody things. I don't know where they're going. I don't know whether they've gone anywhere. I don't know if they're sitting in a warehouse in Taiwan, inside Foxconn or Quanta or Wiriuan or Wistron, who are these ODMs original device manufacturers. So, when Microsoft buy builds a data center, they don't just get the GPUs themselves, put them in their own servers. They buy a server fully built from Quanta or Foxconn, also known as Hon Hai Precision. That's a good company name. Uh, so these companies are probably the ones holding the GPUs if they've even left NVIDIA's warehouses. Because you can on an accounting basis, do something called a transfer of ownership. You can say, sign paper here, these are yours, even though they've never left my warehouse. As long as, and there's nothing illegal about that. Like this happens all the time, like especially with, especially with like very like quick manufacturing. In this case it's peculiar, and I think the most peculiar thing is by this point, I don't know if we've even installed all the GPUs sold in 2024. We don't know where they are, and the capacity is not come online even globally to really cover these sales. It's very unclear what's actually going on. And, I don't know. I just get this feeling at some point something has to break with NVIDIA. Like, they're, I don't think they're doing anything dodgy, other than their GPU somehow making it to China. I think that they are just pre-ordering these bloody things, and people are just saying, yeah, here's my credit card, go nuts or here's a here's a procurement form, will wire you the money. I think with NVIDIA, what it's going to come down to is something will slow at some point, but if any of their customers have the ability to cancel orders, they are quadruple buggered. They are absolutely screwed because if people can just pull back their orders, it's over. I just think that it's really weird that NVIDIA always beats and raises as well. Because if you look around the AI industry, no one else is really making money other than them and the RAM manufacturers and the storage manufacturers. No one like the AI companies don't have a ton of demand that would suggest that they need that compute. Open AI only had 2 gigawatts of compute last year. I'm just really beginning to wonder what the reality is, what the actual fundament of the AI bubble really is, because it doesn't appear to be like masses of revenue other than for NVIDIA, for GPUs that we do not have the physical capacity to install. doing some basic maths, if you've got six months to install a quarter of GPUs, by the time you're installing the last one, it's already two years since since you bought it. So like, what why is anybody buying GPUs right now? If if it takes so long to get them installed, aren't people just kind of buying expensive, expensively built warehouses full of scrap metal that they're going to kind of have to pay to recycle anyway. That is a bloody good question. I mean, there are Blackwell GPUs going online. Like, we know that Blackwell is available. We will see, you can get them on all sorts of different providers. How much is a good question. And the price is already kind of coming down on them, which is a worrying sign. My question is, NVIDIA does a yearly upgrade cycle. So by the time you got your, by the time you get your Blackwells in, you've got Vera Rubin. It's like, uh, I don't want these stinky Blackwell chips no more. Throw them in the bin, I don't need them. I got Vera Rubin, but by the time you, if you buy Vera Rubin when it comes out, uh, I don't want Vera Rubin, I got the new thing. At some point, to your point, yeah, it is kind of silly. The only devil's advocate thing could be is like, oh, they're actually installing them in old data centers. But because of the racks they use, the Oberon racks, you can't install them in old data centers. You either have to gut them or build a new one. Uh, but I've looked around, I've looked around at CoreWeave, CoreScientific, Applied Digital, Equinix, all them. The capacity isn't coming online on the big providers. So is this just 310 megawatt data centers? Like, is this just very small smatterings of them? Where are these things going? At some point something has to break, and the yearly upgrade cycle is going to kill people at some point. I mean, Microsoft made a comment last year that, oh, they're waiting out this cycle or something. Satya Nadella was quite cloak and dagger about it. The cover story with Stargate Abilene was, oh, they didn't want to extend because of Vera Rubin. All of this stinks. It's all, it's quite stinky. There's just, there's something up here because you've got so much money in one bit, but elsewhere not that much. And when you go looking, when you actually really try and find data centers full of these things, it's kind of hard to do, especially at scale. One thing that I've kind of realized that nobody's talking about in in we talk about the shelf life and then it's three years, but then what happens to that card afterwards? Because what is the plan to I presume recycle all of these cards. And then the second part to the question is, what does this kind of mean for companies like CoreWeave who have been using GPUs to back their debt? Because if we're going to have a massive amount of GPUs that may never make it even into the into the racks to be used, then being scrapped, that's going to surely going to really make paying back that debt kind of impossible. I mean, yeah, that that's the thing with CoreWeave. CoreWeave as well is funny because they are, they barely brought 300, 400 megawatts of capacity online last year. Like, they these companies, these NeoClaus, they need to bring on way more capacity than they are. Because that's the only way they make these giant deals. And the way they work, CoreWeave, for example, is they get a big contract from an Open AI or a Meta or what have you, and then they say, hey, bankers, look, I Meta agreed or Open AI agreed to pay me all this money. Lend me money so that I can build a data center so they can pay me. It's why you've got like giant pos to deferred revenue that they can't eat into yet. Except it kind of looks like they're using it to build some data centers, too, which is a bad sign because they're going to need that revenue to run their company. So, yeah, it's, you have to wonder in a year whether you don't just see a crash in the pricing of Blackwell GPUs if indeed all of them even get installed. I don't know what happens. It's so weird. We've never had anything like this in Silicon Valley history. In the dot-com bubble, yeah, we had unused fiber, but fiber is wire in the ground. Fiber lasts for 15, 20 years. Yes, you need expensive interchange racks and such to use it, but you don't need billions of dollars worth, and it doesn't require massive CAPEX investment to make it work. Same deal was during the dot-com bubble, there was a lot of old servers that people were able to use. That still, again, the scale was nothing like this. With CoreWeave, it's and NeoClaus like Nebius and EN Scale. EN Scale, who is apparently building a data center in Luton, uh Luton, London. I've been out of England for too long. They're going to kill me in the comments for that one. Um, but out in Luton, they um they have a data center they're meant to be building for Microsoft. It's the Guardian found out, it's just a pile of scrap metal. It's just, every time you try and look for like a physical thing, they're like, oh, it's just a warehouse. Oh, it's just a thing. In Denton, Texas, where Corey was allegedly building a data center. They just finally built the final bit of it, I think. They claimed it was delayed because of weather, but when you actually went and looked, the weather was awfully dry at the time. It's just all very bizarre. This is meant to be the AI revolution. Why is it that I can't find anything real? Even if we put aside sort of this vague, inevitable sense that AI will have to pay off eventually to make all of this make sense. I mean, aren't we going to be looking at just a huge pile of unused GPUs anyway because they data centers can't be built quick enough to make them useful? I mean, that has to be the case. Because if they're getting digested into the market, so installed at such a snail-like pace, there is, there has to be an overage. Because if the power definitely isn't getting built in time, the data centers aren't getting built in time. There's already a backlog of GPUs waiting to be installed. The question is at what point do people decide that they're bloody stupid to keep doing it? And with the problem in Iran, the whole war in Iran that America's causing for seemingly no reason at this point, or before, it's only going to increase the cost of debt. And we're already kind of seeing this in Japan, the treasury bond rates going up. It's they're worrying about debt costs there. SMBC and MUFG are the two largest data center backers, two Japanese banks. At some point the debt is going to run out, and NVIDIA is not going to have any any anywhere to go. Not going to have any sales. So, it really comes down to, if we go two more quarters and NVIDIA keeps growing, I'm suspicious that something weird is going on. Because the money at this point is just coming without any real output. These data centers aren't getting built, they're, I genuinely think it takes two years minimum to build a data center of anything over like 100 megawatts. It's sure seems to be at least a year, isn't really clear how fast even a 100 megawatt one gets built. At that rate, we're going to be installing Blackwell into 2027. What's the point? And if these things keep selling, is this just NVIDIA looting hyperscalers? Is this this hardware guys conning software guys? Is this just economic hysteria? Are these things actually going anywhere as well? I feel like I'm going around in circles, but there really isn't a good answer. You talk to the boosters about this and they're like, oh, they are actually being installed. Just like complete like, no, you're not the classic school of thinking. But, I don't know. I just have to wonder what happens next. Because the demand isn't there either, so you're going to have these data centers going online with really no one to use them, other than hyperscalers, Open AI, and Anthropic. At some point, though, reality has to happen and someone's going to need to check for actual money. And I mean, the thing I mentioned with Ferramiy America is probably a sign of what you'll see happen. The moment the actual cash stops coming, builders will stop working. Those builders that are actually working. So that will be one of the signs, but I just wonder when NVIDIA is going to stumble. Because they have to because at some point, if they go on another two, three quarters of beating and raising estimates, who are they, like, where are the GPUs going? Because they always sell more. They always sell more and more and more and they're always more expensive to whom, to where? I don't know. These are all the kinds of reasonable questions you'd expect from the finance media, but I don't see them being asked. You mentioned consumer figures, and what about when you factor in that the end customer isn't paying for AI at the rate that it costs, at the very least? I mean, in particular, if we think about Open AI killing Sora and taking away its and walking away from its billion-dollar deal from Disney. It clearly was burning too much money, but it does kind of indicate that there is also not enough people willing to pay enough money for AI to make sense. Unless, of course, you look at the AI startup market, which is heavily funded by the hyperscalers themselves. Yeah, and it's funny because when you really look at I this week's premium I wrote, I actually went and looked for like the revenues of AI companies. I looked all over them again. Every, it's pretty much an industry of people who turn 100 million dollars into a few million dollars, or a couple billion dollars into 100 million dollars. That is the economics of the AI bubble. Most of these companies have somewhere between two and eight million subscribers, which sounds like a lot, but it's pathetically small. I also suspicious of the revenues of Open AI and Anthropic. They're claiming they're making over a billion dollars a month. Really, you, especially Anthropic who this time last year was making like a couple hundred million a month. You suddenly making a billion or two a month so, so suddenly. And no one else within the software stack has seen a lack of sales as a result. That money is just coming out of nowhere. Is it just like a, the software market is just expanded by several echelons? Because when you remove Anthropic and Open AI, no one's really making that much money. Like, nobody else, no one even close. If Anthropic's making 1.1, 1.2 billion dollars a month, if they say. Cursor is making Cursor, like the one of the largest AI coding companies, like the second largest behind Cloud Code, I'd say. They're making what, 110, 120 million dollars a month. No one is close to Anthropic and Open AI, and it's convenient to say, okay, well, maybe it's because they're so good. No, I don't take that either, like Alibaba is not seeing that much growth. They're only losing money. Minimax, who makes who's publicly traded Chinese AI company. I think they made 50 million dollars of revenue and lost over 200 million dollars doing so. Wow, just these businesses are so good. All of the very popular models that aren't Open AI and Anthropic are making piddly amounts of money. They're making absolute tons. Just very suspicious of it. But removing those two, I'm not sure who the customers of the data centers will be. If the idea is that, oh, we will just send to sell this capacity to Open AI, Anthropic or the hyperscalers, your choice is two unsustainable companies that are desperate to go public or hyperscalers that have proved historically to just fire their vendors whenever they want. They have legal teams big enough to do so. I know that Microsoft is doing a $10 to $17 billion deal with a company called Nebius, building a data center out in Vilon, New Jersey. They it's a publicly known information that they have strictures in their contract that if the data center falls behind schedule, they can just cancel the whole thing. So, I'm not sure whether real economy is here. I don't you've got GPUs being sold for data centers that aren't getting built. You've got data centers being built for customers that can't pay for them. You've got an industry claiming it's the biggest, most hugest thing ever. But when you look at the actual user numbers, pathetic. Couple million users here and there. Open AI says 900 million weekly active users. No one else talks in weekly's, Open AI. You only do that kind of thing when you're trying to hide something. They also admitted, Open AI in an academic study that they did last year, that they double count users when they're logged down. So if you're using on a different device, while logged down, counts us an extra user. Sure, it's popular, whatever, but it's being used as a search engine by most. And then the people in the comments and various emails I get will be like, well, AI coding is changing everything. Is it? Like, I don't even mean to be sarcastic. What is it changing? People aren't being laid off, which I'm not saying is a good thing and being replaced by these AI models. You actually have the opposite happening. You have people writing a bunch of code with AI and having to stick around, because you need people to look over it. You need someone to make sure it isn't rubbish. And even then it still breaks stuff like it did with Meta and Amazon, who lost 100,000, 200,000 orders. It's not really obvious why we're doing this other than, unless the goal was to have the most annoying hype cycle of all time. In which case, job done everyone. We've we've done it. We found something more annoying than the metaverse. If we're building all of these data centers for for a demand that isn't there. But then also these data centers are late, and also these data centers aren't going to be having aren't going to be able to have the power for them once they're built. Why, again, why why are people not looking at this and saying, well, that's clearly something we should maybe try and figure out not to spend as much money on. I don't know. I I mean, there I have some theories. My first is that they don't know what else to do. Tech's hyper growth era is over, they don't have any better ideas. CEOs are increasingly disconnected from productivity on every level, they have been for decades. And so these people don't really experience real problems, so they don't really have any real solutions. So all they're able to do as a CEO of a company is hire people, fire people or buy things. And AI data centers allow them to do all of the above. You get to build a data center, you get to buy GPUs, you get to hire people, who are AI experts. You get to fire people who actually do the work. And if you're an executive that's just using like, if you're an executive these days, all you do is go to meetings, go to lunch and talk to yourself. LLMs are amazing for that. But if you do real work, not so much. I think that it's just the final comeuppance of turning the economy over to MBAs, to turning them over to the business idiot elite. who don't actually deal with anything in reality. Blue Owl, apparently, according to the information, took 15 minutes to decide that they were going to fund Stargate Abilene, for example. The money, people think they're geniuses, and when you ask them how they're going to work this out, their answer is we'll work it out. We're just going to get the power, we're just going to build them. The demand will be there. It doesn't, I, if I had to guess, and based on knowing people in private equity, at least, there are just a great many models that can prove anything. You can find a financial analyst deep within a mass a massive asset manager or business development company who can come up with a financial model that will say anything you need it to. You can find a quant who will be like, oh, I'm going to, I'm going to work this out. I'm going to, I'm going to prove that the power will be there, because what they will do is they will point at the power in progress. Progress, sorry, power in progress. They will say, well, look, all of this power's planned. Mate, I can plan anything. I can plan to make a trillion dollars tomorrow. Doesn't mean it's going to happen. But if enough people get in on the same lie, if enough people say, okay, we're all doing this, it becomes much harder not to do it for people with money. They get people from the board saying why aren't you doing AI. You get people in private equity firms saying why aren't we investing in AI. Mythology grows as a result. The whole software is being destroyed by AI story, completely fictitious. Software is being destroyed by bad debt and bad business. But nevertheless, the rationality of the market is gone. Like, that's an obvious one, but the rationality of debt and investment, and even investment within tech is not just gone, but replaced with a fantastical level of thinking. Just a level of like, yeah, well, if we all believe hard enough, this will simply, this will simply fix itself. If we all say the same thing, and we all agree on it, reality will kind of cluster around it. And the reason they think that is that that's how it used to work for tech. It used to be easy for tech. Tech has had it really easy. It used to just be, build more servers, make more money. Get more sales team, make more money. Raise more prices, get more money. All that you had to do is is turn a knob and number go up, because their existing services still had growth to go, there's still growth for Microsoft 365, there's still growth for Microsoft Azure, and Google Cloud and Google Workspace and the like. And these people had it easy. They didn't really have to do anything. All they had to do is go to meetings and be like synergy and all that crap. You know, we're going to grow number up, and then number would go up because they they raised the prices, and they would crowd out competitors. But they all agreed to do something stupid, because they all had it easy. They all had it too easy, and they thought if we do anything, these pigs will buy it. The hogs will slop up whatever we say because they have, look at the Metaverse. The Metaverse, RIP to the Metaverse. They killed it. Mark Zuckerberg took out a shot. Shot it. Like, they put it to the glue farm like boxer in Animal Farm. And everyone clustered around that. Microsoft bought Activision, claiming it was a Metaverse acquisition. You had hundreds of millions of dollars put into Metaverse startups. The only reason it didn't go on longer was there wasn't enough places to park your money. There weren't enough speculative things. You didn't have an easy proof of concept. Large language models are able to do an impression of labor. And thus, you can do an impression of a company, and thus raise money for an impression of an idea. Does the idea work? No. Big tech has actually generally not had to do anything in real life to scale. They haven't had to build anything, they haven't had to invest in infrastructure. The current conversion of hyperscalers, the big tech companies like Meta, Amazon, Microsoft, and Google, they've gone from these asset light cash machines to asset heavy behemoths rolling around in their own filth, ever building construction. And for what? Microsoft has been working on Fairwater, their data centers in Atlanta, Wisconsin for years. And for what? Wow, you're going to spin up hundreds of thousands of GPUs at some point. Okay. Why? To what end? What's the net result? What am I looking forward to? The answer is nothing. Well, I'm looking forward to being proven right. You know, I'm, I'm looking forward to the truth ringing out. But I just think this era is something we're all going to look back on. Kind of like a, was it Roddy Piper in they live. It's going to be impossible to ignore the people who were pro AI in the future. It's going to be impossible to not look at people like, yeah, they'll just build the data centers, they'll be fine. And be like, I don't, I can't trust you anymore. I don't even want to share a cab with you. You're going to try and open the door while it's moving. You're going to try and open the door of a flight to get off before it's landed. I can't trust you, because you thought LLMs were going to control everything, even though they didn't. And that's because LLMs are really good at conning and grifting. They are both as a as an actual technology and as a thing that can resemble something else, and can be extrapolated from. LLMs allow grifting at a far larger scale than productivity. Because think of it like this, I read every day a comment where someone says, well, the coding models are good now. And I mean, in 6 to 12 months, and it's just like, in 6 to 12 months, I will become Spider-Man. I will find a radioactive spider and it will bite me and I will become Spider-Man. Like, we don't, and there are spiders, there's radioactivity. Why are you being an asshole? I'm going to become Spider-Man. That's about as realistic here.
[33:53]And it's sickening, and it's a direct result of a captured media industry, captured business and tech media, and also financiers that have never been part, people should have gone to jail in 2008 for what they did. There should have been real regulations, they didn't regulate derivatives after the Great Financial Crisis. They didn't stop anything. All they did was spank the banks a little bit and say, no, no, no, you can't lend out in crazy ways. But that didn't stop the banks lending to private equity and private credit. 14%, as of middle of last year, of banks large loans went to non-banking institutions, like business development companies like Blue Owl. We're just going to repeat, repeat the wrong the wrong word, because somewhat of a tangent, but this is still not as big as the Great Financial Crisis, and it will not explode in the same way, because there's not trillions of dollars of speculation. But that doesn't mean it won't be horrible, and that doesn't mean that Big Tech is going to look really isn't going to look really different at the end, and I think that is actually a good thing.
[35:19]Well, on that note, Ed Zitron, thanks for coming on.
[35:30]Thanks for having me. If you enjoyed today's episode, please consider liking and subscribing. Also, you can get episodes of the tech report wherever you get your podcasts.
