OpenAI’s IPO is a move of ‘desperation’ | Ed Zitron

Added: 2026-05-22

[00:00:00]What are we talking about? We're basically saying the tech industry will double its revenue in four years. What are we What are we talking about? That's an insane position to hold. That money needs to come from somewhere. If it's debt, if it's venture, if it's whatever, it still needs to come from somewhere.

[00:00:19]And I don't see where it comes from. And it needs to happen. It's not a case where oh, if a quarter of it, if a quarter of it happens, that is not enough.

[00:00:36]On the tech report with me today is writer of Where's Your Edat and the host of the Better Offline podcast, Ed Zitrum. Welcome back as always.

[00:00:42]>> Thank you for having me.

[00:00:43]>> Sam Alman is pushing OpenAI into an IPO as soon as September with a valuation of $1 trillion. Or at least that's what Sam Alman would hope. seems as good a time as any to then talk about the impossible maths behind the AI boom as one FT piece puts it. In fact, the FT piece was written by the managing director of an investment bank and it pretty much hits on what you've been saying for a while and incidentally covers what you've said in your newsletter this week as well.

[00:01:12]>> Yes.

[00:01:13]>> That either Microsoft, Meta, Amazon, and Google need to triple their yearly revenues or the bubble will pop. And I just wanted to ask how realistic that ambition is and then where does this go next if 2030 turns out not to be the magical year that all AI dreams come true.

[00:01:30]>> It's hard to estimate exactly how much AI revenue these companies have because they refuse to tell us. Even Microsoft 37 billion run rate for the specific quarter without telling us which month it referred to because run rate is just month times 12. We don't really know what they're making. But what we do know is mathematically speaking, it's nice to see the FT reaching what I reached about a week later. Actually, several months if you count last year's premium, it is they would have Google would have to make another Google search about $200 billion a year in annual revenue, just a new one, just an AI specific one.

[00:02:07]Microsoft would effectively have to have an Azure or two, so 70 $140 billion. And Amazon would have to I mean they just have to make AWS again which is $128 billion annual revenue. So all they need to do and it's very simple is create another one of their single most profitable and big business units in two years maybe a year. A year would be ideal. And there's just even then though if with even saying those numbers out loud that actually isn't enough.

[00:02:40]There really isn't. They need so much more than that because let's take the AI boosters. Let's see the world through the AI boosters view. So like kind of blue and yellow like a dog. But they they let's say they're correct.

[00:02:54]Even then in four years time there will be no use for those A100s those A100 GPUs. And they've definitely not paid for themselves. So it's kind of like okay that's a wash. Blackwell GPUs by 2030 Blackwell will be replaced with Vera Rubin and at least some like LMFO or whatever comes next. We it's still going to like then these ones are not going to pay off even if they magically created these Google search 365 sized businesses overnight. It's still not really going to pay for itself because they're nowhere near it. There's no, they are absolutely nowhere near creating any kind of AI business that can remotely rival their other ones.

[00:03:34]Putting aside the horrible margins, just on pure revenue, they haven't got it.

[00:03:40]Remember, $37 billion annual run rate for Microsoft. And that's with all of OpenAI's compute, all of Microsoft 365 AI co-pilot and their GitHub co-pilot subscribers who were all heavily subsidized. There is really no even if you think the AI boom is real, there is no sign that this is happening.

[00:04:02]Like that's the biggest thing. Putting aside my skepticism, cynicism, the numbers don't make sense. The numbers are not there. They would have to double, triple their businesses from where they are now. That is not possible. Like that's there are no signs it's happening. that would require them to do genuinely incredible things and would also have to magic up revenue software revenue that doesn't exist right now. I think the total revenue of all software is like 700800 billion a year think thereabouts and but when it comes to TPUs and renting out silicon it can get different but let's say $800 billion they would need that they they would need to create basically an entire new industry that doesn't exist and people will say well there's amazing demand for AI whatever whatever but this is even beyond that they would have to each haveund2 $200 billion AI businesses that paid regularly, reliably, and they would each have to happen. So, call it, let's call it an even 200 billion each.

[00:05:08]So, about $600 billion of brand spanking new AI revenue. I just I there's no I don't think they even have a tenth of that right now other than the 50% of their revenue backlogs from coming from anthropic and open AI. Are we in a position though where hyperscalers basically have to burn the money that they're burning right now in the hopes that AI pays off before they run out of fuel for that fire? Or is it kind of you don't have to be able to outrun a tiger, you just have to be able to outrun the other guy?

[00:05:41]>> I think it's I think they are doing this because they know that at some point Wall Street will curdle and they will have to stop doing cap eggs. So they're cramming it in now. But also, as I've discussed before, 18 to 24 months to build a data center minimum. So, they have to do this now because if they think AI will be big in 2 years, well, they better start building those data centers as soon as humanely possible because you literally can't wait. It's not even because the power of AI is here. It's just if this is what you think your company is doing in two years time, you have to build it. Like, if you were building a city, you'd need apartments and houses in it. So it's I see the logic other than the fact that it doesn't match up with reality on the revenue side. And the thing is again putting aside all my cynicism and skepticism just very practically speaking we are talking about hundreds of billions of dollars of new money like new cash because this would be we're not saying they would replace their current businesses. We are saying pe they would still make oh Christ I think Microsoft makes 300 billion a year. Forgive me if I don't know off the top of my head.

[00:06:51]We're saying another 200 billion on top of that. So people would still have to be paying everything they're paying for Google Workspace or everything they're paying on AWS or Azure and what have you. All of that revenue would have to keep existing and then some. So just we're not just talking about these companies would have to increase revenue. we would be talking about the most dramatic increase in software and compute spending in history and a bunch of new money coming in like just brand new cash either organizations spending more which would still have to come from somewhere or new entities existing to spend more and again that money would have to come from somewhere and then putting my cynic hat back on no just no they're not going to do that that there there's no way they could do that this is not I think people are really they're good at seeing like facts or statements in a vacuum. They can say, "Well, okay, they need to make $1.5 trillion, but that's that's a lot. That's such a big number, I can't even think of it. But don't worry, they'll work that out because they love growing stuff. They're always good at growing stuff." But when you actually look at the practicalities of it, they would have to be growing 50% year-over-year every quarter for like several years for this to even start happening. And that's not really happening anywhere. And then there's the other problem which is especially in Amazon's case, Amazon doesn't have an AI software play. They are entirely an infrastructure company which means that if Amazon can't build the capacity, Amazon can't grow that at that rate. And considering Anthropic is pretty much all of Amazon Web Services AI revenue or at least the vast majority of it, I'm not really sure how they're going to reach these numbers because that's the thing.

[00:08:40]This capex is not really realizing revenue yet. And if it where it is, it's murky and they're hiding it. Generally, when a public company especially has good news, they will share it in the loudest, most aggressive way possible.

[00:08:55]The fact that they're all like, "Oh, I couldn't possibly tell you. Oh, I couldn't. Oh, I'm shy. I'm shy. I'm a three trillion market cap company. I'm too shy to possibly share with you what my actual numbers are. I couldn't poss that's that's a strange thing to do."

[00:09:13]And I don't know, I just don't see how I don't see how any of this is possible.

[00:09:19]And even if you don't hate AI, don't think most viewers of this part of the channel at this point do. Even if you don't hate it, hate AI and you think it's going to be big. How do we get there? Because how where's the money come from? We're talking about new cash.

[00:09:36]We're talking about trillions of it.

[00:09:38]Even if you think, oh, businesses are going to get so much out of AI, blah blah blah blah blah, that's the money's still got to come from somewhere. And software and infrastructure spend is not at that level. We're talking about doubling, tripling, quadrupling the amount of money spent on software and infrastructure unilaterally across the tech industry. I just don't think it's possible. This has made me feel a lot better about my next question, which is pretty much that if if we assume that the demand is there, who if anybody could actually afford to pay the amount needed to make the buildout make sense?

[00:10:14]>> I mean, I don't think anyone can like that's the thing. It's we're even escaping the velocity of AI now. Like, it's not just like, yeah, the answer is anthropic or open AI and they are only able to do that if they fed venture capital money.

[00:10:28]Putting all that aside, the money just like the money will have to come from somewhere. The money will have to come within the tech ecosystem.

[00:10:36]The suggestion would be really like doing some rotten economics here. If you're suggesting that you're suggesting that a bunch of people that currently don't pay for any software or any infrastructure will stop paying for it and you're suggesting that there is hundreds of billions of dollars worth of completely untapped software demand.

[00:10:58]people that are currently not paying for anything or people paying for stuff who would massively ramp up their costs and it needs to happen. It's not a case where oh if a quarter of it if a quarter of it happens that is not enough. This is not something where we can have half measures and I don't see how it happens. Like I genuinely I believe my b my bare thesis.

[00:11:24]I genuinely think AI is going into the toilet. But nevertheless, for me to be wrong, so much new money is going to have to exist at a time when debt is going to be harder to find, at a time when businesses are struggling with higher costs and even the the mighty anthropic, which everybody loves, that even they are in a situation where businesses are flinching. The information's Laura Bratton's had some great stories about this where it's like they're nobody knows how to actually calculate the costs of AI. No one has knows how to measure the ROI. But when even businesses are blowing through their token budgets in a matter of months, their annual token budgets, I mean businesses do not know why they're spending that much because Anthropic doesn't allow granularity on a per user basis of exactly what they're using. And indeed, Anthropic doesn't even provide a service level agreement, which is a standard agreement with basically every software company.

[00:12:23]It's all everything's setting up for failure, but everyone wants to read it as positive.

[00:12:29]And actually, here's a great story for you about like how rickety this is.

[00:12:34]Salesforce. Mark Benov, the Carnival Barker of all Carnival Barkers, said the other day that they're planning to spend $300 million in anthropic tokens this year, but he intends to not do so in the in the future. He's very specifically said, "We don't need to do this in the future."

[00:12:50]That doesn't sound like the setup for a huge hugely successful industry. It sounds like an industry still finding its purpose, its feet, its whatever. And so this is we are too far in to be so vague about how this works out and I truly don't know. Going a little deeper down the rabbit hole, maybe could we start seeing Nvidia rapidly increasing its investment in AI companies just to fuel the the the spend on AI to that I maybe I'm going a little mad there myself trying to explain it, but that's circular financing just to keep the build out going and keep their profits somehow going up even though they're feeding themselves.

[00:13:33]>> That's the opposite of crazy. That's actually happening in their 10Q for this quarter. They I think that their cash position, their actual cash only increased despite selling $81 billion of GPUs. Yeah. Where are they Jensen? Their cash only increased $600 million. And that's because I think it was tens of billions of dollars that they're funneling directly into investments.

[00:13:56]They have, I think, $26 billion uh at the very least of investments that just straight investments this fiscal year. I think I forget when their fiscal year ends, but they have 26 billion through that. And they upped their cloud compute commitments. They're going to spend $30 billion in the next five or six years renting back their GPUs.

[00:14:19]Again, where is the actual demand of in how can Nvidia do that? I mean, I know how they can do it. Sign a piece of paper and send someone money. But if that's the case, if Nvidia can buy that capacity, that suggests there isn't a ton of demand. Now, they'll the boosters will say, "Oh, we'll resell it. Oh, this. Oh, that. Oh, the oh, they'll hand it off to people, what have you." But again, why is Nvidia having to plug the gaps in this supposedly demand satures adopt AI at scale, but given what we've been seeing and talking about with companies like Uber and Zillow burning through tokens like you say months worth and well years worth in months. Do you think that 2026 might actually be the year we start seeing AI budgets tightened instead of expanded?

[00:15:19]>> I think there's a good chance of that because right now we're in the opulence phase. We're seeing I have heard from so many Anthropic customers. I've been speaking to them all week saying that Anthropic is currently just mob bossing everyone. They know that that we're in the hype cycle. So they're just going in saying these are our terms. If you don't like them, screw you. Spoken to two different people who pay Anthropic.

[00:15:40]Anthropic is charging its enterprise customers more for tokens than the publicly available API rates. It's dodgy there. But here's the thing. Laura Brown from the information had this great story about the costs and how people are blowing through their token budgets. And the one thing that people kept saying was, "It's really tough to work out what this costs. It's tough to budget for it.

[00:16:01]It's hard to measure." That is a very bad sign. So, I think we're going to see an acceleration within the revenue of Anthropic and Open AAI on the API side, but especially Anthropic because everyone's goooo for them. And I think then we're going to see it hit some sort of wall because think of it like this.

[00:16:20]You're a business. You're spending millions of dollars on these tokens, but you can't really say why. At some point, you're just spending a lot more money for no reason. And I think businesses are going to come to that conclusion.

[00:16:32]But within these many stories about token burn within organizations, there are always quotes of, "Yeah, we're still working this out. Yeah, we're we don't like these costs." Whenever you see a CFO quoted, they're saying, "Yeah, we we got to pull something back or like this can't this can't last." That is again not a stable industry. when the in the era the advent of cloud computing one's actual cloud storage on the very simple level in cloud compute was a fairly stable and obvious spend it was we have website we have procurement platform attached to it we have this cloud software we pay for and we know why we're paying for it we can measure some ROI but we can say okay we use this ERP system to control where things go financially in the organization we use Dropbox to share stuff we use frame to share stuff. We use Figma for design.

[00:17:24]With Anthropic, it's like, okay, we use this for coding, I guess, and how much does it cost? Well, we don't know yet.

[00:17:32]Like that we we actually don't know.

[00:17:35]That is not something that has long-term viability.

[00:17:38]It is not something we are within the irrational exuberance phase. So, we are seeing people massively overspend right now. Now the idea that you blow through the budget of anything in form imagine take away the word AI here and just think of it like this. You see you have a company and you set a budget for a line item and you blow through it in a single quarter. You are going to immediately start marching into people's rooms and start throwing things. You're going to say that this can't be right that or just we shouldn't be doing this.

[00:18:09]But because AI is the big sexy nasty thing right now, everyone's excited about it. I don't know how long that can really take because within all of these conversations I've had and all these articles I've read, I don't read many people saying why. I don't read many people saying, "Wow, this is so good."

[00:18:30]And the Zillow numbers I reported, which are in my newsletter, those everyone I've talked to there, they're saying like, "Yeah, I don't know why I'm doing this, man. They're just they're telling me to I got I got I got to use agents to build stuff, I guess."

[00:18:43]And across the board when I talk to people in organizations that are having AI forced upon them, what they do is they said, "You've got to use more AI."

[00:18:50]Okay? You got to use it. If you don't use AI, you're going to fall behind and we expect you to use AI. So what they do is they create weird little automations.

[00:18:57]They create like, I don't know, a thing that you can feed a certain kind of document into it and strips out the metadata. very straightforward compsized stuff that you could probably do with Python, but now you've used the very expensive machine to do it because your boss will say if you don't use enough AI, I'll kill you or something. This is I am I'm still even as a skeptic and cynic, I look for this. I am still waiting for a story where someone is excited about this who doesn't sound like they're a member of Heaven's Gate because the people that are excited always speak in these high flu and like this is incredible. It's changed everything. It's always we're shipping something in days that took weeks. What is that something? Can't tell you. I just can't I can't tell you about it. or just they don't. When I see these people saying, "Oh, I've got eight concurrent agents doing this, that, and the other."

[00:19:53]What they building, never see it. The companies claiming that they're, "Oh, we've got all of our code written by AI." Their products aren't better.

[00:20:00]There's not more of them. There's not the only company shipping a ton of new features very quickly is Anthropic. And the the features they're shipping are buggy. Claude Co is still a mess. the most of their products are still a mess.

[00:20:16]But I think they're flooding the zone with them because yeah, I would do the same thing if I was running Anthropic and I wanted to grow my marketing effort. I would just release something new all the time because right now the media will blindly write up anything anthropic sends them. They will say, "Wow, this is going to replace X industry goes here. That's stocks within that industry will drop. Anthropic, it's great for them. Is it good for society?"

[00:20:42]No. Is it attached to reality? God, no.

[00:20:45]But it does it make for a good headline also no. But does it make anthropic happy? Yes. kind of touches on something that uh a few discussions with Natasha Bernal in a few other episodes where and it's echoed in the comments as well that the the the goal of work has shifted from making money for your business or making money for your boss or whoever to using AI to make it seem like what they spent all of this money on is somehow worthwhile and wasn't actually just another software fad that's gone into the workplace and just annoyed everybody. So this is a theory I have called the business idiot thesis which is we are run our economy is dominated by and kind of caters to business idiots manager think to your bosses not your boss specifically of course Isaac but bosses and executives in general how much work do you think a CEO does not very much is the truth work with a lot of them they some work very hard but I don't you don't they're not really doing the day-to-day effort managers in general are no longer expected to do work. They are just people that sit around and tell you to do things. So these people, these do nothings. When they use an LLM, wow, changes their life because LLMs are very good at creating work adjacent stuff. Workshaped stuff, I think is the term. I forget who used it, but it's a good term. So they make this workshaped stuff. And to quote Mobitar, incredible new technist who's up and coming. Uh they're really good at demos.

[00:22:16]They're really good at making to a kind of hasty prototype that sort of works or at least has flaws that someone who doesn't really know what they're talking about can correct. Uh, sorry, can can explain away. They can say, "Well, we'll fix that in the future." Even if that isn't true. This is generally how software is sold. Software is sold with prototypes, with demos. As a result, an LLM can burp those out and somebody who doesn't know better can say, "Wow, if I could do this, imagine what an engineer could do." This naturally creates the pressure on the people below them. But the problem is with the era of the business idiot is we've been in it for a while. It's a postjack Welch of GE thing. So these fake workers, these managers and executive bozos that don't really do anything, business has always been around catering to them, catering to their needs and catering to what makes them feel good. We've seen it many times over with, we saw it with the metaverse. You had metaverse strategies and sil um what's his name? Um the bloke from Service Now in 2022 was saying that he was going to make a metaverse storefront. Mark Zuckerberg changed the entire company's name to Meta to cater to that. You had people doing V a very small amount of people doing VR based Zoom meetings or Horizon Worlds, pardon me, just to make the boss feel better because the boss wanted to feel good.

[00:23:40]LLMs are the most heightened ridiculous version of that in that they made a boss ingratiation machine. The LLM exists to make a a dullarded a dimwit executive go, "Wow, this can do anything because it impressed me between going to lunch and ignoring my emails. I am so impressed by this, everyone must use this. And if they're not using it, if they don't think it's good, I would be wrong somehow. I would be dumb for being excited about this. So, you must all now dance for me." It is a sign of a deep rot within our economy. It is a sign of executive ignorance. There is no executive no publicly traded one at least who talks about AI normally. It's not like you have one guy or woman who is saying you know what this is interesting you can do some stuff with it but you know it's very expensive.

[00:24:37]We've got to moderate these costs.

[00:24:38]Everyone talks about it like they have been on Iawaska. It's very strange and that's because executives, venture capitalists, managers are disconnected from production. They are not part of the economy other than the massive amounts of stuff they exploit within the economy. And as a result, LLMs are amazing to them. If all you do is go to lunch or leave lunch, read your emails, or sit around posting on Twitter, yeah, LLMs probably seem exactly like what work is because you don't do any.

[00:25:11]Getting back to some of the the macroeconomics of of the whole thing, I saw a statistic which explained that AI or tech investment in spending is responsible for something over 90% of America's total economic growth in the financial year of 2025.

[00:25:29]And it kind of raises the question, could hyperscalers actually scale back their capex spending if they want if if they wanted to because a meaningful margin saving cut could trigger a recession which would then hit the wider economy and then undercut their enterprise software revenue anyway. I mean they have to at some point that that's just that's really it because right now the reason that hyperscal is not getting buffed by the markets is because their growth has yet to slow in their other products. Google search is keeping up AWS in general.

[00:26:05]Amazon the store is keeping up.

[00:26:07]Microsoft 365 business and productivity kind of waving but intelligent cloud segment is still doing well. So, Azure and Associated Products, their other businesses are still ripping. And that's because these companies have massive sales operations and they have insane amounts of monopolies. They have massive monopolies. There's just you can't compete with their scale or their sales team or their infra teams. You just can't do it. There is it's why when you look at every other software stock, you're like, why are they not growing with the power of AI? H and everyone says, well, that's because Microsoft, Amazon, and Google have the power of AI within them. No, it's because they already have existing businesses that are doing well. If Google search, if Judge Meta had actually bothered to regulate Google in the antitrust trial, I don't think we'd be having this conversation because I think Google would have pulled back. If Google was afraid of regulators, they would act differently.

[00:27:01]And so, I think that they are still able to cover it up. At some point, they will have to cut capex because the data centers are not getting built. They're being constructed, but they are not being finished. And I think that if people if people don't think about this very clearly, which is don't watch the mouth, watch the hands. Watch the actual money coming in and out of the business. Watch how specific they get with AI, or should I say non-specific. Watch how specific they are about their other business units. They'll tell you exactly what Intelligent Cloud made. They'll tell you exactly what Google Search made. They'll tell you exactly what AWS made. They'll tell you some they'll like give you a riddle for AI revenue. It's like you have to go on a vision quest. You have to talk to the two BS who one tells the truth, one tells a lie from labyrinth. Like you can't get a straight answer out of them.

[00:27:58]And that's something that only happens when nothing is happening or when something bad is happening. And it's still incredible to me that this is considered a fringe opinion. It's I don't know. It's everyone I've ever met in a bad relationship. It's like they'll come up with any possible any possible way to rationalize why things aren't going well. Every possible bit of hope and they're going to keep doing it and as long as things keep going well for the other businesses, they'll be able to keep this up. We need media pressure on it, but I've given up on hoping for that. Going back to the IPO news, the FTP piece I mentioned earlier describes anthropic or OpenAI's IPOs as a major transfer of investment risk essentially from those that have already bought in onto retail investors and pension funds.

[00:28:44]And I know we have spoken about this before. I think we touched on it last week specifically, but do you think that is why Sam Alman is so keen to push for an IPO this year? And then I suppose to build on that, wouldn't that then mean that Alman knows that the underlying maths of the AI bubble is impossible if he's trying to offload this risk?

[00:29:04]>> Well, one second. There we go. Um, anyway, just to answer this question, uh, I think Alman is just doing the end run because he knows for a fact that the magic is dying, that like at this point nothing is really nothing's working out in the way that it should. Because think of it like this. You want to keep a company company private as long as possible to maximize the potential IPO return and you rush for the door if you think you've maxed out on private investment which they all have. I actually think what's an what's anthropic is slightly different and I'll get to that in a second. With OpenAI we are less than a month or we less than a month ago. It's April 26th. The Wall Street Journal had the story saying Sarah Frier CFO of OpenAI does not think they're ready for the scrutiny of the public markets. But don't worry, they're finally to go public in September. This is just desperation. It They probably knew SpaceX was coming. They knew They knew the IPO was coming. They knew the S1 would drop because it was the same day they announced it. They knew that was going to happen. So, they said, "We've got to get ahead of this. We got to get our investors excited."

[00:30:11]And that's the thing. I think investors are going to be excited. But this feels desperate. It feels rushed. And it feels like something that you only do when the the other sources of funding are running out because I don't know with Anthropic I think they should be going public too.

[00:30:31]Like if if they have better economics than OpenAI, why are they holding off as well? Do they think that they can get a 1.5 or$2 trillion IPO?

[00:30:40]It's unclear.

[00:30:42]But I will say I hope OpenAI files to go public. I think it's time we see inside the guts of these companies because the SpaceX IPO has revealed they lose billions of dollars doing this. Like they're having they're spending $2 billion on gas turbines. They're standing up the most messy data centers of all time. When you actually see the underlying economics, you see that these are not good companies. So open air probably just wants to get there first.

[00:31:09]They think they think that if Anthropic goes first that they won't be able to go public at all, especially if their economics are worse. It's hard to tell.

[00:31:17]What I do know is I don't know if I ran OpenAI, I would either rush as quickly as possible or never go public because the moment they go public, it will be tougher to raise money. It will actually be tougher because you'll have to use your real economics. Open AAI may also be under pressure from from uh SoftBank.

[00:31:34]There was a big story in Bloomberg saying that open the OpenAI's sorry SoftBank's stake in OpenAI actually is kind of like it's worrying people at the organization. They're worrying that they're overly reliant upon this. So ultimately they may have investor pressure or they may have tried to raise another round or talked about raising one and found no takers or alternatively they may just want to dump it. Orman doesn't have any stock, but I imagine his investors are saying, "Come on. Come on, Sammy. Let's get that out of the door. Let's get this done." I don't know. I want to read that S1 so bad.

[00:32:11]Anyone who brings me that S1, you're a legend for for life. I can't wait to read it. I think that once we finally see the economics of this company, everyone will finally shut up about it.

[00:32:22]on the topic of investment pressure and just finally a Janus Henderson survey of a thousand wealthy US investors found that over twothirds of them at 67% were worried about the AI bobble popping this year and kind of following on from what we were just saying how much is that kind of market sentiment a threat to the the status quo and are these the kind of investors that we're talking about here or are they unlikely to be the first ones to jump ship >> so within the greater hysteria of the AI eye bubble, everything makes sense right up until it doesn't. Because if you think about how insane this is, big tech's planning to spend what 900 800 900 billion in capex this year, 1 trillion next year. This all makes sense right up until it doesn't.

[00:33:09]The kayfabe is being kept up that AI will work out unless they stop. Nvidia what $81 billion in the quarterly revenue selling GPUs. I think 60s something percent of their revenue came from three customers, which means if any of those three customers decide to stop spending, Nvidia's revenue drops. And for Nvidia's revenue not to drop, people will need to stop spending hundreds of billions of dollars on GPUs that don't exist yet, and they can't install them.

[00:33:38]The reason that these investment types are worried about this is it's obviously insane. It's obviously bonkers. It is crazy what they are doing. A trillion dollars in capex in a single year. What are you doing? What is the That is more I think than the combined capex of like 2012 to 2022. It's complet I'm actually pretty sure that's correct. It's completely insane. And some of that capex is being driven by Meta who has no AI story. Every time Meta talks about how they're making money off of AI, again, it's the Riddler. It's well yes if you look we have a 4% raise in this month within Instagram kind of with dogs and cats but not parrots and it's just again you or I regular people when we have to prove ourselves we have to prove ourselves with an alarming clarity you go for a mortgage they look straight at the P&L of your life when it comes to AI it's like oh don't worry about spend as much money as you need we got your back get as much debt as you need that's not a problem you're building AI That's a good thing. And I think that I think it's despicable. And I think it's why regular people find AI, one of the reasons regular people find AI so grotesque. But also, keeping this inflated is costing so much money, so much more money than has ever been spent on anything. There has to be a payoff at some point. Maybe it runs for another 6 months. Maybe it's 12 months. But the reason they're afraid of it popping is because it's only kept inflated by belief. It is a cult-like belief that nothing will turn into something or a very small amount will turn into a larger amount of revenue than has ever been made soon like two to three to four years maybe.

[00:35:32]What are we talking about? We're basically saying the tech industry will double its revenue in four years. What are we What are we talking about? That's an insane position to hold. It's a truly insane position because that money needs to come from somewhere. If it's debt, if it's venture, if it's whatever, it still needs to come from somewhere. And I don't see where it comes from. I have no idea. And I don't think anybody else does either. So they're just hoping and praying that the world somehow works out.

[00:36:05]>> Well, Ed Zitron, thanks for taking the time.

[00:36:07]>> Thank you. If you enjoy today's episode and you want to hear more of the tech report, please consider liking and subscribing. Also, you can get episodes of the Tech Report wherever you get your podcasts.