Bursting the AI Bubble with Ed Zitron

look, you know, I don't know when I'm going to leave this job. I don't know when I'm going to stay in this job. I don't know how these things are going to do. But when I do, you should replace me with Indeed. If you need to hire, you got to say to yourself, this is a job for Indeed sponsored jobs. And, you know, you put your job site on other job sites, like for instance, a telephone pole with one of them papers. You know what you do? You put a paper up and you cut seven slots. Generally, this is done for learning guitar. But you can also do it with jobs. And then whoever walks by that telephone pole will rip one and call you. Now, that used to be the best way to hire people until Indeed. But Indeed sponsored jobs will get you quality candidates when you need them most. You're not going to have to go telephone pole searching. Spend less time searching and more time actually interviewing candidates who check all your boxes. Less stress, less time, more results. When you need the right person to cut through the chaos, this is a job for Indeed sponsored jobs. Listeners of this show will get a $75 sponsored job credit to help get your job the premium status it deserves at Indeed.com slash podcast. Just go to Indeed.com slash podcast right now and support our show by saying you heard about Indeed on this podcast. Indeed.com slash podcast. Terms and conditions apply. At Harmony Bank, we believe that you're a part of the community. We believe that better happens together. We're a locally owned bank built on relationships and personally committed to your success. Our community bankers are ready to make quick decisions. No big bank bureaucracy, just a custom approach to help you thrive. From business loans to mortgages and everyday banking needs, Harmony Bank provides a secure, modern experience with a touch of Texas hospitality. We're Harmony Bank, where better happens together. Welcome to the weekly show podcast. My name is Jon Stewart. I'm the host of the weekly show podcast. It is Tuesday, October 27th. And excuse me, I'm going to stand back from the mic for just one moment. Can you hear the people sing, singing a song? It's the election. This is the last time that we will speak before the election. And I just want to urge everybody out there to please, please, for the love of God, get out there and vote and express your political opinion. And I'd like to frame it as a however you'll vote, but I don't mean that. I mean it as please send a message to this reckless shit show and of administration that enough is enough, that we have had it, that it is time for someone to step up and provide, uh i believe the framers referred to them as checks and balances on this uh utterly imperious I believe, the right thing to do. administration so that is that is the spiel for today we are not talking about the election today we will be talking about other doomsday scenarios uh financial ones in particular about ai i know we've talked a little bit on the program about catastrophic outcomes of AI in terms of bad actors or in terms of the AI itself turning on us. This deals more with the real world implications of AI economically and financially. And we're delighted to have a guest today to talk about that. So let's just get to it. Ladies and gentlemen, we're delighted to be joined by Ed Zitron. He's the CEO of Easy Primary Research. He's the writer of Where's Your Ed At newsletter, but more importantly, he has a great understanding of, I think, the economics of AI. And Ed, welcome. Thanks for having me. It's an absolute pleasure. I want to ask you this, and i apologize in advance this is going to be slightly remedial spine in in some regard uh so i wanted you to talk about you know we've heard a lot about ai in terms of doomsday scenarios uh bad actor gets a hold of it builds their own biohazard or uh the agents break out and hack into all of the sensitive data and they or water systems things like that or the reverse it all works, and we're reduced to NPCs in a world that we have no agency in. I want to talk to you about the sort of the more economic imbalances of the AI world. Does that make sense? Absolutely. All right. Can you give us, to start with, kind of a little bit of the playing field of what is the ai economic ecosystem chips software data set what what are we dealing with when we talk about the ai economy so at the top of the stock market largest company in the stock market is nvidia they make something called a gpu technically means graphics processing unit this isn't for graphics it's because they're really good at parallel processing so basically doing one thing a bunch of times nevertheless you need those chips to do two things one to train an ai model and to to run inference which is just creating the outputs when you say creating outputs you mean so uh when you put in a question the answer that comes to you is the output yes exactly that i see okay and then it gets more complex with bigger bigger models and bigger tasks but nevertheless that's the basics so these chips are the gpus and they run the entire ecosystem yes so and nvidia up until the last few years had basically had an effective monopoly on doing this because they have this thing called cuda which is just underlying software libraries and ways of running software on graphics processors that this meant that they were they are effectively the company behind the ai boom now they sell these gpus which are very expensive they're $50,000, $60,000 a piece, usually buy them in racks that are millions of dollars each, and you need thousands of them, hundreds of thousands in some cases, to do these big data centers. But NVIDIA, pretty much the only vendor selling them, there's technically AMD, technically Broadcom has AI chips too, but NVIDIA is the top dog. Their customers are pretty much the largest companies in the world. Microsoft, Google, Meta, Amazon, Oracle, SpaceX, now, I guess. And they have a very small customer base, but that customer base has so much money that they've been able to sink $30 to $40 to $50 billion a quarter into these AI chips and into AI data centers. Now, the problem is that these things keep getting more expensive and also take forever to install. But for the most part, NVIDIA has got large because everybody has said AI is going to be the next big thing. We need to buy as many of these as possible. Not really much proof yet that that's actually the case, or at least anything close to a justification for the trillion plus dollars that have been invested in these GPUs and the data centers for them to go into. I can kind of stop there. There are various levels to fan out from there. Right. But so the general idea is you have this one company, NVIDIA. Now, does NVIDIA, are they supplying it to everyone? Are we including China in that equation? Or are we saying that's merely the United States and maybe other sort of AI startup areas? So technically, China has a company called Huawei, which makes their own chips. But NVIDIA GPUs have been making it over there for years. There's a big story at the end of last year from Bloomberg that there's this company called Megaspeed, just called a NeoCloud. It just means a company that buys chips and installs them and rents them out, where they appeared to be allegedly a front for a chinese data center operation super micro an ai server company was also had one of their co-founders literally using a hair dryer to get the serial numbers off to ship it was like it's like weapons here's the thing this era is so grim and soulless and joyless it's nice to have some goofy hijinks for once like something i think it is so nvidia chips are making over there but they have far less and china has had to be a little more deft with how they're building these things but for the most part nvidia is selling gpus to everyone everywhere throughout the world so nvidia basically controls everything this means that nvidia is incredibly rich they're printing money it means that nvidia has now become the largest supplier to microsoft google amazon meta spacex the problem is because these data centers are not actually generating enough revenue or really any revenue that they don't none of these companies other than spacex disclose their actual ai revenues everyone's had to start take on debt they've had to lay people off meta's doing a 95 billion dollar share sale oracle had to do a 50 billion dollar one uh facing a downgrade potentially all because despite all of the big promises of ai the economic payoff is not that torsten slog head of uh chief economist that apollo said the other there has been no payoff from AI. Well, that gets to the heart of kind of the, maybe the absurd contradiction in all this is that, so you have this ecosystem built around these NVIDIA chips that cost millions of dollars and these, and what it does is it creates, I guess, compute. Everybody's looking for compute, but compute is very expensive. he's looking for compute, but compute is very expensive. Apparently, and you tell me if this is incorrect, but the people that use Compute, the customers, don't pay anywhere near how much the Compute costs to build and run. In other words, you're buying something and you're getting it at ridiculous discounts and the company is taking on that excess cost. To an extent. So this is actually, the thing is, you're both wrong and we don't know, but also right at the same time. I will take that. No, no, no, no. I will take that as a win. Let me explain what I mean. We don't actually know how much it costs to run these things. We have no clarity. We don't know how much it costs. We have a lot of estimates. A lot of AI people online have come up with their guesses, but we truly don't know. We have these company called Silicon Data does an indices of how much you can get per hour for a GPU. And these are always going up and people say, wow. It must be, they must be making the money back. We actually truly don't know the operational costs, but also even those indexes of these prices don't really reflect what people are paying long-term because they're mostly based off of spot prices, which just means what it, what it would cost to rent this right now versus if I buy it for a year. Let's step back a second then, and maybe, so we've got a little bit of the lay of the land to who the players are. How then does Anthropic or Meta, or any of those make money? On their AI, how does open AI generate revenue? Okay. So two ways, one, there are subscriptions. So you pay Glammy Samuel, when you say I'm going to pay you 20 bucks a month, and I can use open AI's models for, and there are rate limits that kind of obfuscate, but you can use them a certain amount. They have a hundred dollars and a $200 a month subscription. Same with Anthropic. So you pay that and you can use a certain amount. Now, Anthropic makes more money on the second one. What? Which is the API, which just means connecting your thing to Anthropic's models. So your startup, your coding harness, and you pay at the meter. So you pay on a per million token basis. Go, go, go, stop, stop, and I apologize. No, it's okay. No, please, please. What is a token? Because I hear that a lot, and I'm not quite sure, and it's, and a per million token, how many tokens does it take? And what is the name of your cat? And for those of you who are listening on the radio. This is Howl. And he is a big, beautiful, annoying, just like daddy. Okay, so. All right, go ahead. So a token is just three quarters of a word. It just, it is when a model spits out an output. Let's say you say, give me a limerick, and it will spit out, and it will be a certain amount of tokens to generate it. So when you put in the stuff saying, make me a limerick, that's an input token. Okay. And when it generates the output, that's the output tokens. And is that calculated on the amount of compute that an input would require to get a result? No, it's calculated based on what it's doing and what it's putting out. The compute side, the how much the GPUs cost to run, what this, we truly don't know. Like we have estimates from all over the shop, but when it comes to especially like Anthropic and OpenAI, who are the main AI labs, who are just the ones that make the most. popular models, we don't have any insight into how expensive their operations are. We have some documents. I got OpenAI's audited financials earlier in the year, but we don't have, Hey, it costs when we run one NVIDIA B200 chip, that's how much this costs an hour. And the problem is, is that it's bordering on impossible to actually estimate because we don't have insight into their infrastructure. We don't even really know how much compute they have and we can. Go into this in a bit, I guess, but the hyperscalers, the Amazons, Googles, Metas, Microsofts, and so on. And those are the ones that are trying to build all the data centers so that they can increase their compute. But if, if we don't know how, why are they all in this race then to create more compute? Are they maxed out or are the inputs that are coming in so much greater than what they're able to export? Okay, so that's actually a very good and important question. Oh God, thank you. No, no, it really, no, because, no, because John, you're being self-effacing about not knowing this stuff. We actually, a lot of this stuff is very obfuscated. Okay. Purposefully so, I would imagine. Yes. On purpose. Okay. So it's not so much the largeness of the outputs, it's that there, it takes a long time to build AI data centers, way longer. Each one of these things is the single most ambitious infrastructure project. In history. Each data center. Pretty much. So open AI's data center, the Oracle has been building it for two and a half years in Abilene, Texas. It is, I think about 998,000 square feet. You'll be, and then it's going to, and it's going to take about 1.2 gigawatts of power, which is basically the power of a small city, but in a thousandth of a space. So I use the city of Bristol in England as an example, it's basically the power of the city of Bristol in a one, in a. Basically several large warehouses. We, we've got, we've got Americans right now scrambling to find out where Bristol is and what it looks like. Yeah, I do the most exciting thing to happen to Bristol. Sorry, Bristol listeners. But so the hyperscalers are building this capacity because they see demand, but really most of that demand is coming from open AI and Anthropic who have this voracious appetite for AI compute to the point that because they have so many customers and so much attention. Mm-hmm. Are basically buying up everything. I estimate that 80% or more of AI compute spend is from those two companies, but hyperscalers have all got deals with just these two companies. Yes, they have some other customers, but when you remove Anthropic and open AI, I think that's like $22 billion, which sounds like a lot until you realize Microsoft alone has spent over a quarter of a trillion. Microsoft has spent a quarter of a trillion on, on hyperscaling. On, on capital expenditures since 2022. So capital expenditures are just. Investments in assets that are something you will use for the future in the company to make the company more money, allegedly. Now, are they, when you say that they're spending money as customers of open AI and Anthropic, isn't Meta also in the game and aren't they, or, or they're a minor player in all this? So that's the funny thing. I don't know how people forget about the Metaverse, why they're called Meta, but Meta is spent over a quarter of a trillion. I think it is at this point, probably more on AI, building these data centers, building out AI compute, but they don't really, like they have their own AI models, but Meta doesn't have an AI strategy. Meta has built their own AI models. They have this Muse app. The little assistant. That's recently popped up. That is very unprofitable. They don't make much money. I know, right? You look, we love this. And this is the thing. All of these companies are doing this because they're out of hyper growth ideas. They don't have a new Google search. They don't have a next Facebook. So they all went because they're all hired, old, boring monopolists. They go, wait, Microsoft did this chat GPT thing. We should all buy these chips, more chips now, big, big. And the problem is they put so much money into it that they can't really walk it back. Even though most of their customer base is Anthropic and open AI. And without them, it's kind of. Piddly amounts of revenue. Folks, I know, I know things seem a little, a little grim right now, but I'm telling you, man, there is a new project on news that is here to save us ground news. It's our old pal ground news. They are here to help you deal with the news based relations. relentless insanity the algorithmic uh driven bowl of sadness and despair that pours forth through every outlet and orifice that you can find your way to. It is madness. But Ground News, they take every article about the same news story from all the outlets all over the world, and they put them in one place, and they tell you where it's coming from. You can see how these different organizations and algorithms manipulate your information. They can show you how reliable the source is, who is funding the source. It's a hell of a service that they provide. Subscribe for 40% off unlimited access by scanning the QR code on screen or go to ground.news.js. Make sure you use our link so they know we sent you. Again, that's ground.news.js. I mean, all you hear about is this is a revolution on the scale of, I don't know, the cotton gin or the railroad or, you know, and some would say, you know, the iPhone and the Internet and how that changed everything. They, they certainly feel as though they have discovered the future, whether or not the rest of us want to live in that future is sort of a separate question, but. but so they're are they they're making a bet that this is transformative yes that they think and here's i must be clear these people are not super sophisticated they may be rich but they are not sophisticated november 2022 rolls around chat gpt comes out literally we are just out where we have gone through what four rate hikes that year maybe more everything's really expensive the tech industry had this massive boom era in 2021 right and some inflation is very high we're coming out of the pandemic supply chains are depressed revenues are down across the board nvidia's revenue is flat and then back in 2019 uh nvidia had bought a company called melanox in israel which allowed you to just use high-speed networking to connect a bunch of servers inventing the modern ai cluster now okay 2022 comes around chat gpt and because remember these hyperscalers are not invest inventive they steal from each other they they copy each other it's all they can do they all see chat gpt and they go oh that looks like a new thing what if we put a bunch of money into it and that's what they've done since they they weren't now to to be fair to them when chat gpt came out it was a sensation yes it did ripple like the effects of the introduction of that technology uh people around the world suddenly started uh asking their computer to write breakup letters uh as though they were drake like people weren't necessarily using it for anything pretty children can't use that that's right but it was they were absolutely it was adopted very quickly without any real sense of what it might be useful for so it was it wasn't as though it was an entirely fabricated product it wasn't as though it was an entirely fabricated product it wasn't as though it was an entirely fabricated product yes i agree however yeah it did go viral the media took it the next steps i would say the media ran away with it but also the hyperscalers put all the money into it and so there was this natural belief that okay this is real because there's a bunch of money behind it i'm not doubting that large language models are interesting and that there was something there i've just been saying this whole time that the actual economic outputs and actual technological things they are claiming it can do aren't really matching up with what they've said so anyway so the the idea is is is and that the user fees are never going to match the infrastructure sort of the idea of if netflix came out and uh they could only charge 50 cents and that's the only thing that people would pay but it's going to cost them a trillion dollars to make every show and so it's just not it will never match up that the money that's coming in will never match up with how much it's going to cost you to create and maintain the technology necessary for ai is is that somewhat in the neighborhood yes so open ai and anthropic are really interesting companies economically speaking because they lose billions of dollars anthropics s1 just came out they lose less billions than open ai s1 is it's the filing you make when you go public so you file it when you go public so you file it when you go public you have to file an s1 with the sec and they are about to go public so they have to file an s1 with the sec okay exactly and what is in the s1 well what's funny is what's isn't in it all of the stuff for the latest quarter is blank so all of the numbers that you really need to understand what anthropic's like blank they've only got up until june 30th now to be clear they shared many of these numbers with investors at an investor day two weeks ago so oh well screw the retail investor i guess right now what's the difference between the two of them interesting is the company scraped by with a light profit in q2 but that was because elon musk discounted their compute at spacex for two months so whatever and we don't their their compute is owned by spacex some of it is this all just a circle jerk what i don't understand so they're all building everybody's building data centers yes and they're all renting space at each other's different rates so just to be clear the ai labs the anthropics and open ai they rent compute they rent compute from space well open ai doesn't but anthropic pays they're going to be as of december they're going to be paying 3.1 billion dollars a month a month to elon musk so so when did elon musk build all the compute i had no idea that he was even in this and why wouldn't anthropic or open ai be building their own compute well that's the problem these companies lose billions of dollars which makes banks really uneasy when it comes to lending you money so what they do is they have their hyperscale pay picks so oracle larry ellison everyone's favorite larry ellison is building about 400 billion dollars worth of ai data centers for one customer sam altman clammy sammy himself is going to apparently pay him 300 billion dollars over five years now does open ai have the money no does oracle have the money no but don't worry according to td cohen nvidia is going to be helping oracle out by leasing space and then subleasing it to them nvidia by the way is the g capital of this all they invest in all the companies they invested in open ai spacex anthropic core weave which is a neo cloud so a company that just builds data centers and sells ai compute and by the way takes out debt using the chips using that it's magical so they they take out with money they don't have they're leveraging so they're highly leveraged that's core weave yes on their own build out and yet they are fighting not to pay taxes on this value that they are saying they have but don't have it would happen yep is that the magic trick there's so many magic tricks that's the thing because the ai labs is it's funny people are always talking about the open ais and anthropics as in oh well they're building their own stuff no they're not anthropic has this is serious 430 billion dollars worth of non-cancellable commitments with amazon google microsoft and broadcom broadcom that ai chip company yes uncancellable uncancellables you cannot cancel these commitments and what are these commitments that they've made so for amazon google microsoft they've made a lot of commitments they've made commitments to rent data center capacity hundreds of billions of dollars worth and when i say non-cancellable it's something called a take or pay arrangement which means you are like you're going to use this pay you're not going to yeah you got to pay regardless okay so anthropic has it's funny the way i'll describe the s1 is it's a description of a company that doesn't exist anymore because it was before all of these commitments begun and anthropic now has these things probably starting next year that are going to cost it 75 to 100 billion dollars a year within the next year or two and this is regardless of whether its revenue grows this is a company that only loses money what are their project what are their how are they justifying these commitments where where are they seeing then the growth in all this how do they justify putting that much money into a build out and usage they believe that they will grow to hundreds of billions of dollars of revenue in the next few years because if they don't they're not going to be able to pay this their operating expenses will be 100 billion plus in short order microsoft so a very profitable 330 billion dollar a year company has operating expenses of like 150 billion dollars so anthropic just believes it will be the size of microsoft soon you may be wondering where the proof is it isn't there i am wondering that that was one thing i was wondering yeah the s1 doesn't explain that anthropic just has these crazy projections their valuation comes down to them making 200 billion dollars a year in revenue in two years 200 billion and what is their revenue now uh probably in the region of 30 to 50 billion a year max it we don't have the fourth quarter oh this is the this is the worst episode of shark tank ever yeah but what's crazy is you tell people this and they will say well it's different this time these are the smartest people the smartest guys in the room if you will you know enron uh and these people are so smart ai is the next big thing and they're going to be the next big thing and they're going to be the next big thing but here's the really scary part so according to uh it's a payment and credit card company called ramp 80 percent of open ai and anthropics revenues come from one percent of their customers and most of those customers are venture-backed ai startups that run their services at a massive loss okay okay okay we i want to roll this one back because this one all right so my understanding of all of this is is probably steeped in the television business again netflix streaming i i'm thinking of it in terms of customers who pay a fee to utilize a service that is necessary for them to run their business so we're and so we're talking about the hundred dollar two hundred dollar a month whatever they might charge a company to use their compute you're saying that's only 20 percent of their 80% of their revenue are actually actually other ai startups and those ai startups what are they are they're not paying a user fee what are they paying for so remember i mentioned the api earlier so connecting your stuff to the models yes what is api i'm gonna forget the term but it just means the the way that you connect your service to another service so you're running there's a ai legal service called harvey they use the api so that their software runs using open ai or anthropics models jillian just popped into the chat application programming interface they're gonna have my guts for garters for getting that one uh software that talks to each other so they have to connect to uh the the big brain or their ai startup does not function yes and then there's the other problem which is on these companies and when you you you you , when you use say perplexity which is an ai search company and you pay them 20 bucks a month you can run up a token spend of like 60 bucks or 70 bucks or 80 bucks a month so they're running at a loss exactly and what's crazy is is that everyone says wow anthropics growth open ai's growth it's so impressive how is it so impressive the answer is is that they have a bunch of venture-backed startups that run their services or at a loss and have users who say hey you're i want to use the latest models the latest more expensive models the latest models as well the prices may seem lower or similar but they burn more tokens and they actually output tokens so they create the thing that it gives you those are more expensive so everyone is crediting the growth of these companies to some sort of diverse revenue model when a large chunk of that revenue is based on basically the ninja borrower and they're not going to be able to make a lot of money out of it is the assumption here because it's a and we'll deal with this in the realm of anthropics valuation which you know people are saying they're going to have an ipo they're going to raise 1.5 trillion or maybe even two trillion yeah they're going to be valued at that and raise like 100 billion they're going to be valued at that and and raise all that money but what that does is they've exhausted the private venture money and now what they're saying is oh you know where you know who we can get the money from now people are going to be like oh you know who we can get the money from now people are going to be like oh you know who we can get the money from now people yep just fucking people just dupes like they're going to and what's crazy about this as well is back in 2012 there's this thing called the jobs act push through so that it was meant to open up capital markets more people but it had a subtle change that meant that private companies it used to be that private companies over a certain size would have to report their actual revenues they'd have to file like a public company so it forced companies to go public uh much more than that lower valuations so google went public i think their valuation was 23 billion there are three four trillion now i think microsoft was like 700 million they're worth trillions now amazon a few hundred million anthropic is going out at 1.5 to 2 trillion dollars this is exit liquidity what is exit liquidity tell me tell me what that is oh i'm getting that don't worry so we are sold this we are sold this idea that we are investing in the future with these companies that we're investing in the next google the next meta the next what have you when actually what we are is the dumping ground for venture capitalists that have invested in the company we allow them to exit the company by selling their shares that they bought at a much lower valuation for a massive payoff and in return we get a company that what is anthropic even if anthropic was to grow to the size of nvidia it would be what 3x 4x which is a good return i guess but compared to what you used to get from these startups but also it's not going to do that this is a company with negative a bit so a negative basically earnings before tax taxes depreciation all right so let's take let me let me push on it in a kind of uh little devil's advocate so sure it's a similar story though i remember when google came out and they had these wild valuations and they were getting crazy amounts of venture capital and people were saying how do they make money it's you're building this they didn't call it compute then but it was the network you had to build out you know fiber optic networks that were taking over some of the dot-com bubble build out and but how are they going to make any money and over time though it did turn into a profitable business whether it was through advertising or the selling of data or other ways that they made money either by sucking up other companies or things like that are we missing that same story with ai are we being too too skeptical about how they can turn this into a profitable business so let's take amazon amazon is the one people like to choose amazon web services which is their cloud platform one of the most profitable businesses in history incredible stuff founded in 2003 i think it was available to the public in 2006 profitable in 2015 amazon's combined capital expenditures in that period to build out cloud capacity it was $29.7 billion adjusted for inflation for some context over how long a period 12 years between 2003 and 2015 and that's not per year that's total that's the total capital expenditures $30 billion and anthropic raised $30 billion in february they raised $65 billion in may amazon spend i think $45 billion in the last quarter and here's the thing as well amazon at that time was a much smaller company so if you invested in amazon at that point you were having an incredible return on investment if you held on it paid off in spades and also the difference is is that anthropic and open ai do not have assets these companies don't have data centers they're not buying data centers they're leasing data centers they're leasing chips but that's the part that i understand why wouldn't they want to own their own infrastructure how would they afford him or is it impossible but then what are they so all the money they're raising is just on leasing is just on agreements that they're making with other companies yep that's exactly him hey folks i know most people look at me and they think to themselves well that that's a fashion play right there that dude he knows what's up i'm surprised he's even got time to do the podcast knowing that uh he should be in milan sitting with anna wintour but actually uh you know what i am comfortable you might be thinking well sure if you wear the same t-shirt as last week but do you know where you can find clothing that fits into this quince quince has got all the false staples you actually live in denim boots leather jackets the unbelievable soft what do they call in there 100 percent mongolian cashmere sweaters starting at like 60 bucks come on i'm not talking about you know des moines cashmere mongolian cashmere so what makes quince different they work directly with ethical factories they cut out the middleman uh you get top quality materials and craftsmanship at half the price of other premium brands find your next fall favorites at quince download the quince app for app exclusive offers or go to quince.com get free shipping on your order in 365 day returns now available in canada and the uk too and when quince asks where you heard about them let them know it was the weekly show best way to support the show and here's the other thing somebody's gonna fail and if one of them fails if you're talking about these hyperscalers spending 60 billion dollars a quarter or whatever you're talking about that's going to create i mean if you think about the mortgage crisis in in 2008 that was i think they said it was something like seven percent of mortgages went underwater and that destroyed the entire derivatives market around it so it's not going to take that big a hit to ripple through this industry and create a disastrous domino effect yeah and what sucks is as well is the american stock market is basically held up on like a toothpick of ai oh i think only it's the max it's only those seven companies are driving the entirety of the gains of the stock market yeah like uh 45 i think at the value of the nasdaq 100 is in the magnificent seven and by the way the valuation of the magnificent seven is largely based on the idea that ai will be the biggest most strongest thing and grow forever, NVIDIA's continued valuation, I think $2.15 a share, as I think about this, is based on the idea that NVIDIA will sell more and more GPUs, basically selling a trillion dollars worth of them in two years. And we're at the point now where outside of Microsoft, who has just offered buyouts to 27% of its staff to keep bankrolling AI, Meta, Google, Amazon, cashflow negative, all cashflow negative having to take on debt, they can no longer buy these GPUs out of cashflow. So the NEO clouds never could afford to. So they've had to take on onerous debt, core weave, the biggest of them and the worst of them. I think that if they were to raise debt today, it'd be 13, 14, 15%, basically a personal loan. So this is all based on debt. So yeah, the moment anything bad truly happens, that debt stops getting issued. They stop buying those GPUs. Do you have a sense, Ed, that these businesses, we always think of it in terms of leverage, how over leveraged are these businesses? And again, going back to 2008, when they found out that banks were leveraging themselves at 30 to one or 35 to one and endangering the very sort of foundation of their businesses, if shit hit the fan, what is the kind of, have they calculated the sort of leveraging of debt of these companies and these build-outs and lease agreements? So when it comes to the big ones, like Microsoft, Google, Amazon, Meta, they're not going to do anything. They're not going to do anything. Nvidia isn't going to go bankrupt. Nvidia as the chip seller is going to probably take a haircut, but the big hyperscalers are going to be fine, but in trouble because when they stop doing this, everyone's going to ask why they did it. CoreWeave has $30, $40 billion worth of debt. Morgan Stanley said they're going to need $100 billion more. What is CoreWeave? What is that? CoreWeave is a, it's called a NeoCloud. It is a company that raises debt, buys GPUs, builds data centers. And by the way, CoreWeave is a, it's called a NeoCloud. It is a company that you'll be shocked to hear that their customers are Microsoft for open AI, Google for open AI, Anthropic, Meta. Yeah. You know what this reminds me of? I hate to even bring this up. You probably don't. There used to be a show called Beavis and Butthead. Do you remember the show? Yes, of course. So there's an episode of Beavis and Butthead where they're selling, you know, kids used to get those, you'd get a tin of chocolate, you'd have 10 candy bars, and you would sell them to raise money for like the pep squad or any of that kind of stuff. And so, uh, one of them, had a dollar. I think Beavis had a dollar and he said to Butthead, give me one of those chocolates. And he gave him the dollar and then he ate the chocolate. And then, uh, Butthead said to Beavis, you give me one of the, and they just passed the dollar back and forth. Yeah. And by the end of it, there's no more chocolate and they still only have $1. Yep. That's basically it. Except NVIDIA gets the real dollars. Yes. NVIDIA gets so many real dollars. Well, and Ellison gets the real dollars. The landlords are still getting, no. Okay. Talk to me there. Oracle is in a weird position. Okay. Our favorite, our favorite leathery old man, Larry Ellison is, um. How dare you, sir? That's my new, that's my new daddy. All of the Ellison fans screaming in the audience. No. So Oracle is in a peculiar situation because their public relations is phenomenal because people say Oracle is doing great. Oracle's revenue has been flat for 15 years, adjusted for inflation. The only part of the company growing is AI compute, which is destroying their margins because it costs so much to buy it and takes so long to install. And then Larry had this wonderful idea where, Hey, this open AI company that loses billions of dollars, I should build endless compute for them. So September last year, they announced this $300 billion, five year long contract with open AI and the analysts gooned. They went completely nuts. They were saying, Oh my God, this is so amazing. Now Oracle stocks tanking because everyone went, wait, can you afford this? Wait, can open AI afford? What happens if open AI dies? Who pays for this? Oh, no one. There's no one else. There's literally no one who can afford this. Nobody. Oh, that's bad. But even then the market hasn't fully ingested it. So what this on a very grand scheme, what this is, is AI to the greediest of us represents something. And to all of these big companies, they believe based on what I believe is kind of a horribly misread demand signal that there is insatiable demand for compute. When what's really happening is two very large companies are buying all of them. And because these companies are run by the big ones are run by people without vision, they say, well, these companies are growing real fast now, which means they will grow even bigger and faster in the future. It is that simple. It is not an elegant or D it may be complex in the execution, but the logic is extremely childish. So they're all, they're all building an endless, compute for a demand that they're not sure the public will even be asking for that. There'll be a plateau. Yep. Like I'm assuming like, like everything else. And I hate to keep going back to streaming, but it's kind of the only business that I mildly understand, but there are plateaus in subscribers. And so what happened with that business is everything is chasing penetration. So you want to get, into as many homes as you possibly can, but at a certain point you run out of homes or certainly run out of homes that give a shit. So then you have to double back and create other methods of revenue within that plateau. But it sounds like they're making an assumption on an, a non plateau system. So they haven't demonstrated the demand or the usage or the profitability, but they're betting that they will. Yes. And what's funny is it's not just demand they need. It's it like people could want AI, but most people don't spend that much on AI. Most people won't spend money on it. The vast majority. Well, you said mostly it's going to be other AI startup companies that need to use their interfaces. So they're, that's where they, that's where they believe their big money will be. And even more worrying. I saw a few people on Twitter talking about this. There are AI startups now who are renting their own GPUs and they're renting them. And this is the scariest thing they've said. They've said, don't worry. There's so much demand for compute that if we don't need this, we'll just sell it to someone else. That's right. It'll be excess compute that will be allowed to, but it sounds like they're just going to be selling it amongst each other. It's kind of like when in the, in 2008, when people were saying, well, if I can't refinance my house, I'll just sell it. When, when does this bill become due? Do you know, like how much this profligate spending you talked about, you said they had spending commitments. Did you say $60 billion or am I imagining that? Oh, it's much bigger. Anthropic has in the next seven to 10 years, $430 billion worth of non-cancelable commitments. OpenAI has projected to spend $750 billion by 2030. So, and I imagine that they also have servicing that debt. There there's probably, I don't know what interest rates will be in that moment. And I guess you can't project that, but I imagine they also have to service that debt on the commitments that they've made. Well, that's the thing. It isn't even debt. Well, technically, so Anthropic technically has $161 billion worth of leases with, with Broadcom, which does carry interest, but these commitments are otherwise just contracts. I see. So the debt is being carried by the hyperscalers, by the people that are building. Exactly. For everyone else. This infrastructure. So they're not, they're tenants and everyone else is a landlord that's building out. And those people will have to service the debt. If AI companies like OpenAI can't fulfill their service leases. Exactly. And then there is the other thing of, even if OpenAI and Anthropic had the demand, even if they needed it, there's the other problem of how slowly these things are being built. And again, we get back to how little we know about this capacity. So Microsoft spent $265 billion on AI, on AI GPUs and data centers since 2022. They have said in the last three quarters, each quarter, they're like, we brought on a gigawatt of data center capacity. Bloomberg reported, and I reported with the Guardian as well, that actually Microsoft only has two gigawatts of data center capacity. So they're like, we brought on a gigawatt of data capacity. So they're like, we brought on a gigawatt of AI compute capacity. Operational with the rest of it sitting in warehouses. Amazon business insider just reported has billions of dollars of chips that they either can't plug in or well, they, they either can't plug in because there's not enough power or they just haven't built a data center yet. So on top of all of these other problems, we have this situation where hyperscalers are buying way more GPUs than they need driving up the price of memory and thus the price of consumer electronics for everyone. And on top of that, they're lying kind of about how much capacity they have. They are trying to make everyone believe they're bringing gigawatts online when a data center can do many things. The data centers that people hate are not the ones that do streaming. These AI data centers are completely different beasts to anything that runs anything else. These things could all disappear tomorrow and the internet would function the mostly the same way because they only exist to do AI. Well, the AI thing was always, it, it, it wasn't a necessary, development. in any way no you know i i always felt like ai is just like google that google that's on coke like when you google something you say like hey man is there a pizza place near me and it tells you but when you say to ai they're like oh my god that's such a great question i love pizza pizza so good did you know it started in greece like it just tells you shit and you're just like shut up and sometimes it goes yeah demarcos is the best demarcos doesn't exist yeah that's right hey man i need all your water i need all your water yeah it doesn't exist and it makes shit up yeah so let me so so here i want to now like take a step back and give people a sense of like here are the vulnerability points up so like you've got a bunch of vulnerability points one is these ai startups that are using the apis will uh fail to the extent that they lose eight you know enough of their 80 of revenue that their revenue crashes a little bit yeah the anthropic and open ai um i don't know if that's a growth problem or just a chip surplus because they've got too many chips and people aren't building them fast enough let me correct on on nvidia because nvidia is a very peculiar case so nvidia doesn't care whether you install them jensen huang gets paid regardless indeed it appears that business insider guy could eugene kim had a great story about this amazon has been doing an early rack receipt program which means that they have to take chips before they're ready which is very clearly nvidia forcing them so that nvidia can bump its revenue and they're not going to be able to buy them because they're not going to be able to buy revenues but that that that's not sustainable that that they'll have a revenue growth slow it isn't and on top of that they have another problem you're talking about weak points i think it's deutsche bank says that 50 to 60 percent of nvidia's revenue comes from hyperscalers so like a handful of companies most of them requiring debt so if the debt don't come they not buy chip and chips by the way are getting more and more expensive so it's not just nvidia needs to sell more chips are getting more and more expensive so really simple this is going to sound technical but it's pretty simple so the reason that every consumer electronics your phone your laptop everything is more expensive is because there are basically three companies that make all the world's ram micron sk hynix and samsung now that makes all the world's ram meaning memory yes so random access memory so that it's in everything okay now nvidia's gpus which they're selling millions of regardless of where they go use something called high bandwidth memory it's a special kind of ram and it takes up more space to build it on the fabricator leaving less space for the useful ram that everyone else needs which the ai servers also need so long story short because they've sold so many of these chips the triopoly of ram companies have said we're just jacking up the price because they can because they can and so all of these gpus being sold in many cases months if not years in advance being built for no reason are pushing up the cost of every single consumer electronic so that's just another fun thing with nvidia that's just another thing that we can blame the ai boom on so what is in in your mind is there an event horizon on this is there a moment uh a singularity of economic uh catastrophics that are they racing against the clock or are we underestimating their monopolistic creativity in the very high stakes circle jerk like are we not being are you looking at it too rationally that was my mistake a few years ago i i really was like that the the the nuance of my analysis now is that i've given up on childish things of rationality so you used to have more of a a calendar on this you thought this was a ticking time bomb and it's going off at such and such a moment you don't believe that anymore no but the things that can destabilize this the most obvious one is these venture-backed startups so they have to raise money tens hundreds of millions of dollars every three months so this is an interest rate game if interest rates go up anymore boy that could create an enormous issue in the vc world well it's that and also just venture capital will run out of money venture capital has been on an insane losing streak since 20 yeah since 2017 there have been thousands of these things called zombie unicorns companies valued at one billion dollars i know it's great term one billion dollars but they can't raise another round because if they raise the round their valuation would go down because they're not worth as much because they were overvalued and um or they can't get acquired because they're worth they're worth on paper too much so venture capital since 2017 the average vc fund has failed to return more than like one dollars and thirty cents for every dollar invested so you literally would have been better than the s&p 500 on top of this through between 2018 and 2022 a bunch of software startups got vastly overvalued so there's a bunch of stuff they can't sell i mean that exuberance is part of any of these technological you know breakthroughs but there is another layer to that okay which was that i think 30 of private equity take privates in that period were for software companies so you have private equity with thousands of companies they can't sell either and now venture capital has one really big mistake which is they're saying ai is going to be the biggest most hugest thing ever even though most of these companies are just rappers around open ai and anthropics tech but because they need so much money they need to but they need to get money from vc every few months every time that happens because everyone's excited the valuation goes up and up and up and up and up right and up and up and up there is a company the reason i mentioned harvey is they've raised like over a billion dollars in the last year or so billion dollars jesus and they're worth 15.5 billion dollars 15.5 billion dollars for a company with no real ip it's mostly just software cognition's another one another one i mean they're just a business it's very unclear exactly what they do they have a coding agent some other stuff they've raised billions of dollars and they're worth 48 billion dollars there's no ipo is there a scenario where this works where you know look we were i remember pets.com you know when we went through the internet startup and everybody's like and and everybody rushed those were all the zombie unicorns back in that day and vc came in and all these people and anybody that had an internet startup suddenly you know it was everybody was we work and they all went in there and a lot of those collapsed out like everything else and but there was a solid infrastructure left standing that was worth something yeah is there a world where because this it's clearly worth something it may not be worth the every company is a two trillion dollar you know a behemoth but it's clearly worth something how does this get solidified what's the scenario where this works out less catastrophically so the problem is is the dot-com bubble was the actual software part of that was teeny tiny and comparison these companies were valued at hundreds of millions of dollars which was a lot yeah and now would probably be a billion dollars given yeah the amount of time that's with inflation whoever knows but nevertheless most of these companies raised fractions of the amount even adjusted for inflation that one of these ai startups does perplexity an ai search startup with basically no mo of any kind in a dying customer base is raising billions from nvidia by the way from nvidia at a 30 billion dollar valuation there is no power for these companies even with a pivot they're not going to be worth even a shred of that and the venture capitalists are going to get washed out boohoo they deserve it venture capital needs they need an apocalypse event because they no longer invest in valuable companies they invest in financialization and private equity the same oh yeah it's all the same financialization and here's the thing blackstone we can go into but blackstone may actually face an existential event too but really i'll get to that in a second because the data center part's really important i know it's the blackstone but it's the blackstone blackstone we can go into but blackstone may actually face an existential event too but really i'll get to that in a second because the data center part's really important i know it's the best part that's the best part ai data centers are useful for one thing ai now the comparison that people make with the with the dot-com era is the telecommunications companies the dark fiber there's all this fiber in the ground the expensive part was the fiber in the ground once you had the fiber most of the work was done for you companies die use the fiber easy peasy not with the ai data centers if they're incomplete then they're going to need just as much money in three or four or ten years to finish as they did back then and then there are the opex the operating expenses the electricity is going to be just as expensive if not more so in a few years this isn't getting cheaper and everyone wants to say oh it's like the dot-com bubble because you know after that which included like a massive drawdown and nasdaq was horrible and hundreds of thousands of people lost their jobs and it was very horrible and i think people are too blase about that but nevertheless there is no dot-com bubble recovery here there is no there are no alternate uses for gpus and the demand for ai gpus as side of anthropic and open ai is minuscule so while it could exist after this we have according to sunshine climate i think it was like 190 gigawatts worth of capacity in planning even if a fraction of that happens that's still 30 40 50 gigawatts and we're talking hundreds of billions of dollars worth of demand just to make the bill to get the bills paid and these data centers are paid with debt i don't see how that turns into something else after this and people People really want it to. And I think that people choose the dot-com bubble because, first of all, it allows them to believe the system functions correctly, that finance and tech can't be wrong. Right, that it can correct itself. Exactly. However, it also allows them to say, oh, this wasn't a big waste. And here's the thing. While the hyperscalers might be able to repurpose this stuff, we are building these data centers everywhere, and most of them are barely complete. All of these GPUs sitting in storage are not useful for other things. We can turn those buildings into malls, you know, the country- Laser tag arenas. It could always use more malls. I don't think that'll ever go away. Yeah, I mean, but even dead malls in places are another fascinating story of commerce. No, I know. I'm kidding. Oh, today's ad. Today's ad is bowl and branch. You know, I remember when they first- When they first got together, like Dow and Jones, when bowl was doing fine and branch, you know, both, listen, they both were doing the best they could. But when they got together and formed bowl and branch, and thank God that a company that sounds so pleasant makes good stuff. They got great sheets, 100% organic cotton, good airflow. Competing, man, it's all about the airflow. You're always thinking. You're too hot, too cold in the thing. You get airflow, give you a little softness without that trapped heat. Babush, I say. Make pillows, waffle blankets, comforters, temperature regulating comforters. 94%, by the way, of the bowl and branch customers say their sheets actually get softer with every single wash. Which I'm going to tell my son who was in college to encourage him to wash his sheets. Tell him, they're bowl and branch. Dude, just wash them. They'll get softer. Nah, shouldn't have outed him like that. Point is this. Upgrade your sleep with bowl and branch. Get 20% off your first order plus free shipping at bowlandbranch.com slash TWS with code TWS. That's bowl and branch. B-O-L-L and branch.com slash TWS code TWS to unlock 20% off exclusion supply. Do they believe their own bullshit? Like, does Sam Altman, if you pin them down, you know, the Amadeus, do they believe that it isn't going to be about customers, that there will be such a productivity surge in the economy? It will replace so much labor for the corporations that these giant bets will pay off. Like, what is their. And I don't know if you know this, and I don't know if you may think, like, well, they're deluding themselves. What is their version of success in all this? How does this carousel stop? So, Dario Amadei is deeply embedded with this group called the rationalists. And rationalist does not mean rational belief. It's the rational belief that superintelligence will kill us all. I think that. So, that's just one of the people influencing him. Do you ever hear about METR or any of these organizations? They are a religious cult. It's insane that they are poisoned. It's everywhere. It's truly. This guy called Adam Becker, you should talk to about this. He did a book called More Everything Forever about it and many other things. But basically, it's a billionaire cult that believes that superintelligence will destroy us all. And the large. And yet, they're all building it out. Yes, because don't you want to get killed by American tech? But they're also, to be clear, the promises are so far. So far from what all of this actually is. But Amadei's argument, and I actually think he may believe this, but I'd also believe that he's a cynical grifter, is that, oh, this will cure diseases. There's no proof of that. None. Oh, it will. Autonomous. It will run businesses. No real proof of that. That still doesn't make it profitable. And by the way, they keep saying this, and it drives me crazy. They say, what is the benefit of this? It's going to cure climate. It's going to cure cancer. And you're like, oh, well, what are we using it for now? Uh, finding undocumented people, targeting, uh, weapons, uh, companies are using it to use, uh, uh, pricing that's dynamic so that they can charge. Blown up schools in Iran. Right. So it's all. They keep saying there is this utopian world that we're going to get to with this incredible intelligence. You're like, oh, what are we using it now? And they go, well, mostly exploitation. It's mostly exploitation and extraction. And business stuff. Right. And the thing is, the biggest. And I would argue only consistent market is coding. Like, it's the one thing it's fairly good at, and even then, fairly good at still involves it being really great sometimes in the complete more on others, and hopefully you can spot. But that wouldn't justify the cost of it, because that whole coding market's not worth that. Here's the thing I'd say, and people get really mad, the booster types, the AI fans hate this, but, wow, OpenAI claims to have solved an ancient math problem. Wow, great. Have you asked the. Have you asked the same machine how to make a profitable business? Because apparently that math's problem is just too big for. You can solve all of these historic problems. Have you tried to do that? Have you gone on, like, Claude and said, how can I make AI profitable? That's like, is Claude bullshitting itself? Well, the answer is always, they'll just make it more efficient. Here's a nightmare scenario for you. Do you remember when DeepSea came out, and everybody was like, wait a. So, okay, this is another thing, and I'm sorry to keep bouncing around. I'm like this, and I appreciate your patience. No, please, please. So, generally, these technologies get cheaper as they get more efficient. They get just like everything. You know, the progress is always like, how long does it take to make a baked potato? An hour. I can do it in 10 minutes. Like, it's supposed to get cheaper and faster and smaller, and that's kind of how these technological progressions are supposed to go. This doesn't seem to be the case with AI. It gets more expensive. It needs more of our. It needs more of our resources. It needs more electricity and water and space and cooling and everything else. But is it possible that what we don't know about what China is developing, as they did with DeepSea, is a model that is pretty much as good, but runs on much less compute and power and water? I mean, in theory, but. But not really. Well, I mean, also, Minimax and Zipu are two Chinese. Two Chinese AI startups that are public and. Zipu? Zipu. Z-H-I-P-U. All right. But nevertheless, they're public, and they are also unprofitable. So. So we're all going down together. China's going down. We're going down. Doesn't fucking matter. I mean, going down. I must be clear. I don't think this is great financial crisis level bad, but that doesn't mean it won't be very bad. Well, here's the thing. It'll be different, which could be. How do you measure bad? But no area of the economy, though, is not at some level integrated now into this project. And everyone is invested in the market now. 20% of the stock market's volume is retail investors. Everyone has exposure. When I say not as bad, I'm even hesitant to say that. I just think it will be a different kind of bad, because great financial crisis was massive amounts of speculation and terribly done CLOs that had other bits and pieces. Right. And it fell apart. Not necessarily because every asset was broken, but because everyone lost confidence in the system. Almost no assets were broken. We literally, the world fell apart because 7% of people's mortgages were underwater. Which is. That was the entirety of the fucking crisis. And the thing is with this is it will be different because it's one of the biggest problems could be entirely symbolic. It is the loss of. It is the loss of confidence in the AI trade and the Magnificent Seven. And that is a large chunk of people's S&P 500 and NASDAQ 100 and NASDAQ Composite holdings. So should we shut up if the whole idea is none of it's real, it's only a confidence game? Should we just be like, it's going to be great? No. In fact, I wish we'd have said this earlier, because the thing is Microsoft, Google, Amazon, Meta do not disclose their AI revenues. How is that legal? How is that even possible? Because you don't. They report their gap earnings, so generally accepted accounting principles. They report legally, but they don't have to. They don't have to do anything. They don't have to do shit. So they don't have to tell you how much money they are making. And because the media, the finance and business and tech media goes easy on them and has for years, no one really says boo. And when you tell people, well, they go, well, look, and the way that people justify it is they go, well, look, their revenues are going up. Fun fact, Meta increased the cost of ads. Google increased the cost of ad. Microsoft increased the cost of everything. Amazon increased the cost of, I mean, things in Amazon Web Services. They turned the fees up on their sellers. They increased revenue through other things. But because people are saying, wow, these are the smartest guys ever. They wouldn't spend hundreds of billions of dollars on nothing. So that growth must be coming from AI. And the growth they are seeing from AI, by the way, is from Anthropic and OpenAI. OpenAI.com. By spending money on compute, and you'll never guess who funds OpenAI and Anthropic. That's right. Microsoft, Google, and Amazon. And also NVIDIA and Broadcom and Micron. No more chocolate and they only have $1. It really is. That is the story of all of it. And what sucks is as well is all of this could have been avoided if we'd been a little more sensible and actually had real financial regulations, but we don't. And what sucks is Anthropic will probably go public. Anthropic will go public and the retail investors. No question. And they'll get dumped. They'll get this company at $1.5 trillion, so a little bit under Meta's market cap. And Anthropic has intentionally in their S1 not given us a timetable of when their big payments kick in. They don't tell us what their cash position was in Q3 2026, even though they have that information. And they've done it because in the next six months, Anthropic is going to have tens of billions of dollars of new compute costs out of nowhere. And retail investors will have no idea because the media, ooh, ooh, look at Anthropic. Look, Anthropic had a one profitable quarter because they got a discount from Elon Musk. Oh, that's fine. Don't worry about it. Except I do worry about it because there are going to be people that buy into this stuff and they're going to be people that think because they've been lied to, they're buying the next Google. But what I don't understand is where their credibility comes from in the first place. And you see it like, let's just take Meta as an example, you know, and social media as an example. For a year. They went down to Capitol Hill and said, I'm telling you, we're doing this right. We put in all kinds of restrictions and things to protect people and nothing will go wrong. And yet every time you turn around, they're paying $20 billion here for harm that they caused in because they didn't do the safeguards that they said they were going to do or that they didn't, you know, whistleblowers will come out and say they knew that these products were harming people and they didn't give a fuck. They just kept going. And so we're doing this. We're making the same mistake we made with the financial crisis with is don't worry about the crimes you're committing. If you just give us a cut of it, we'll be fine. We won't make you change your behavior. If you could just throw us 5 billion, we'll just call it square. I think that it's and this is grim to say, I think the media is a large part of it. No question. Since 2023, the media was saying that. I was doing stuff. It couldn't do. There were stories in the times about a task rabbit being manipulated by an open AI model that didn't happen. It was a guy asking an open AI model for what to say to a task rabbit to manipulate it. What's a task rabbit? Is that a task rabbit? Someone you hire? It's someone you hire to hang up pictures. It's not. And in this example, it was to solve a web thing. It was just so fucking stupid. But nevertheless, the media was saying posting every few weeks. That's Sam Altman. Very scared of his own stuff. Yes, and then exuberance. And then when the AI, when the hyperscalers were spending all this money, I was going to people in the media and saying, hey, why aren't they saying what they're making? And people would say, it's the early days. It's just like the dotcom bubble, the same old tired tropes instead of scrutinizing the people who were spending the money. And the thing is, I've definitely been wrong on timing. I thought this thing would wrap up in 2025. Couldn't have been more wrong on that one. But nevertheless, the underlying economic problems are the same as before. And they say then they haven't corrected themselves. It just hasn't proven fatal. And it just continues to go on. So if we're looking at because so now I'm just trying to sort of this new technology and find a way to categorize where the advantages and the threats are in terms of the threats. You have sort of that, as you were saying, you know, the Amadei who is so afraid of his own technology and believes that ultimately it will kill us. But at least. We will be killed by American technology and not Chinese technology. There's the area that I find is the one that is most likely, which is bad actor using really powerful technology to create something dangerous and that that creates all kinds of problems. And then there's this third area, which is financial house of cards. Where does your kind of lie awake worrying about these things? Where do you place or the fourth category, which is it all works and we live in a utopia with high universal income and nobody has to work and we've all lost agency. So where where where do you place in that continuum? So I think that there is a real. So these models going out and hacking, they're not going rogue. They don't have an accord. They don't have thoughts. Nevertheless, we have open AI, anthropic, and I think Google Meta as well, running dangerous automating. It's software at scale using hundreds of billions of dollars of infrastructure, and we have no legal, we should, Sam Altman should be in prison. I think they, I think that they, here's the thing. An open AI model hacked an Australian government website. If a regular person did that, they would be in prison, go to jail. That's right. Why is Sam Altman not at least at the police station? Why is Trump not on television going? Clemmy Sammy is terrible. Matt, you would love that, but by the way, that's pretty. Oh, yeah. I've been working, working on it up for hours, practicing. Now I'm, I like it, but I think that that is the most obvious problem right now. We have this unrestrained, unregulated tech being run on massive infrastructure and in the tests, they're doing real damage. And now we have open source. They had no reason to train the models to do this stuff. Why are, why are they beta testing things in the public square? And when they say it breaks out of the sandbox, why you're, you're absolutely right there. And everybody just kind of shrugs their shoulders and up until this explodes, that is going to be an escalation. It's going to be an escalating problem where basically hard forcing compute to break things by bad actors using open AI's models mostly or open source models that have been distilled. So trained off of open AI and anthropics models. My big thing is the financial thing, because I mentioned earlier Blackstone. Blackstone has over $200 billion of data center infrastructure. Who funds Blackstone? CalPERS, the California pension system, pension funds, individual investors. Wow. Okay. And so this data center debt and Blackstone is in core weave. They're in open AI. They're in anthropic air truck. They have bits everywhere. They're in Broadcom's deal with anthropic. They have tens of billions of dollars of deals, partially funded by pension funds. Everyone is exposed to this. Apollo merged with Athene, a life insurance and pension fund company back in 2022. Athene is backstopping part of one of anthropic's loans. I, the good news is, is these pension systems. They're huge, so billions would be bad, but they'll still keep paying pensioners. But I don't know, because of the private part of private credit, I don't know how widespread this is. That scares me. Look, here's what, here's what ultimately happens. And I think we've learned this is that the federal reserve ends up just attaching a hose to whatever is bleeding and pumping it full of quantitative easing. Like they're just going to print money and pump it to bail out because of these things. Yeah, that's what I meant. That's what I meant. But they end up doing it. Yeah. They're going to bail out Blackstone. And also, they're not going to bail out OpenAI and Anthropic. And even if they did, you're going to bail out the $430 billion of commitments from Anthropic. And on top of all of that, you can't bail out a stock market. It's too much. That is, you can't bail out the value of NVIDIA, unless the government is going to give NVIDIA $1.2 trillion a year. If they are, Microsoft, Google, Amazon, their growth futures are based on Anthropic and OpenAI spending. Is the government going to prop that up? No. So, maybe they bail out little pieces of it, but you can't bail out a stock market that's drunk on growth. It's interesting. It used to be, you got to make yourself too big to fail. They've actually made themselves too big to not fail because we can't bail them out. And also, OpenAI is too small to pull into component parts because OpenAI is fascinating in that I think Dario Amadei might believe something. I think he's got the religious cultists around him. I think Sam Altman's just the most cynical guy ever. Right. I think he just goes and says whatever the last. Whatever he thinks the last guy would like to hear. Well, you see that in his interview. He says different things in different places and you just, you don't know what any of it means. And he'll occasionally say something insane like with Jimmy Fallon where he said, I couldn't imagine raising a child without chat GPT. Mate. But the thing is with Altman is he is doing what they're also all following the Silicon Valley playbook, which is sign as many big deals, make number go up so big, so huge. And then thing happens. It gets cheaper, right? Except as we've established that. That isn't happening. And the problem that I imagine they have is with OpenAI and Anthropic is that the demands on them from the world are you will grow to be the next Microsoft, Google, Amazon plus plus, which requires you to act like it, which requires you to sign all these contracts. And because we are surrounded by what I call business idiots, we are in the era of the business idiot, the kind of the, it's kind of like the, the final bosses of neoliberalism, these guys who only care about growth. The Satya Nadella's Hoctan. and so on these guys who everyone claims are the smartest guys in the room and ron and what they do is they go well these people are promising me all this money how amazing would it be if i got paid that money are they thinking can they pay no open ai is the fastest growing company in the world anthropics fastest growing company they grow so shell game and the thing is everyone the media regular people investors buys into it because it's such a part it's a powerful confidence game it's they all believe the same thing we hope and because they all believe it no one wants to stop and there's always that tinge of truth which is this is incredible technology it's like nothing we've ever seen before i'm not quite sure how to use it but i'm sure they'll figure it out but that's the thing though that's the crazy thing about this though even in describing this thing four years trillion plus dollars in what's it good for code professionalizing your emails i don't need that i have fingers like right and it and by the way and i've seen a lot of people who have done this and i've seen a lot of people who have done this and i've seen it i don't really use it much but the times i have i realize like oh i'm getting dumber i think just during this session yeah yeah you you're you're having concepts served to you rather than understanding the concepts and you don't even right i need to be connected to the world that i live in i just have to that's just the way it works uh but ed uh uh it's it's absolutely man thank you thank you for for helping to uh really give me an articulation of the shape of what this uh financial shell game really looks like and what the component parts are uh i found this really really uh helpful frightening i'm not gonna lie but but really helpful to understand it so i really appreciate your patience and and taking the time uh ed zitron ceo of easy primary research uh he's got the where's your ed at uh newsletter and uh man man knows his shit uh thank you sir thanks so much at harmony bank we believe that better happens together we're a locally owned bank built on relationships and personally committed to your success our community bankers are ready to make quick decisions no big bank bureaucracy just a custom approach to help you thrive from business loans to mortgages and everyday banking needs harmony bank provides a secure modern experience with a touch of texas hospitality we're harmony bank where better happens together you know when you talk to somebody and they have recall he knows a lot of stuff what my head i started thinking like do we even need chat gbt or can i just call this guy a lot of facts and figures the level of detail on a variety of issues you know when you read a book sometimes and you have to go back and go like what does vituperative what is what is like and he's well the you know if the s1 doesn't go through and then you've got they 39 and it wasn't like when i know things even then i generalize it like it's around 40 but the 39.6 billion dollars in s1 when they filed the thing but they didn't show you and i was just like holy shit and he's on pacific coast time he's up early and he looked quite fresh i like how he said data centers i gotta tell you everything sounds more authoritative when delivered in just a hint of that he didn't have he wasn't full-on cockney but we had the london accent just a hint of that did this motherfucker go to eaton and even if not he could have you know like he was nearby so let me ask you guys so i am in that data in that ai doom loop a little bit of not knowing where the threat is and also not seeing the promise of the upside of like we're going to cure all these things for you but all they're using it for are these exploitative surveillance uh state issues and weapons issues as we laid out the various threat matrix that goes along here is the banality of just the economics of it that does that move to the front of the class as far as threats where you guys come from coming out on this that's kind of where i've been oh really okay just because it seems really unsustainable and we've all seen you know number go up and to and it won't it's like you know completely irrational right and it seems like everybody's just paying each other like you the beavis and butthead um comparison that you made everyone's just paying each other the same dollars i'm a very sophisticated analyst round and round so you've been on you've been on bubble watch a while because i've been on somebody's gonna make a bio like that's where i've been well that's like the um that's like the back of your mind that's always there you know where you're like right well and then also you know we're all gonna die oh and somebody might right britney same with you uh no i think the economics of this was a surprise to me like i haven't been following this so closely and i i know that we use the term circle jerk often but this is circular financing jerk right yeah but this i apologize to all the all of our parents out there we do tend to go with circle jerk quite a bit it always applies it's crazy it works uh but i was like holy shit this is the ultimate one like that was very surprising to me right to hear the level of which i didn't realize the big surprise to me there was that open ai and anthropic they're just tenants i did not know that their business model and that larry ellison is really and elon musk they're the landlords yeah but they're all on amazon and meta right but they all live in the same town and then at other times it probably flips and those people are giving money to the other people it's it's a wild little house of cards they got going i mean this morning nvidia announced they were going to invest like three billion in meta and it's like i guess meta would just be giving that money back to you at some point and way more than that back to you basically i'll give you three billion dollars and you're going to make a 15 billion dollar chip purchase yeah along with data center leases and everything else it's it's a it's a wild little game but i i really appreciate it uh britney what what are the people thinking about are are we are we getting a lot of questions on midterms where are we at we've actually first one yeah john we've got one week left if the democrats win big do you think trump and co will just january 6th it oh god i don't you mean like just not seat the representatives and senators it's it's honestly And as we've seen over the past episodes, I am not a man known for my lighthearted vision of the future. I mean, I am optimistic at my core about people, but people exist in systems that are designed by, I guess we'll call them assholes. It's hard for me. It may be a failure of my imagination, as January 6th was too, by the way. I remember watching it in utter disbelief. But it is hard for me to imagine, and I do believe, and I'm not saying we're Paris, and I don't mean that in the burn-it-all-down way, but I mean it in the, I do think there is a point where people will go to the streets and those elections, and I don't mean it in a January 6th way. I mean it in a we-are-protesting-an-authoritative-power-grab kind of a way. I do think these are the fulcrum points. Mm-hmm. Where a society decides it's enough already, and we're no longer going to play this game. But I'm clinging to the idea that somebody there learned a lesson from that terrible, terrible day and will function as a governor on those impulses, but God knows I could be very fucking naive. Yeah. Come on, Jillian, you got something cooking in there. Come on. I mean, I'm not, I think, you know, everybody says it, and it's true, we have to come out in, like, overwhelming numbers, because you can fudge the numbers a bit when it's, you know, 10,000 votes. You can fudge, what did he ask for in Georgia? Like, 13,000. 13,000 votes. You know, those are numbers where you, they can start fucking around. But we all have to show up, so I don't, I think even they are not praising that. No, and it would have to be egregious. Yeah. I mean, whatever. It's like when they say, it's why the worst question that I ever see newscasters ask as it comes to these election cycles is they always say, you know, does she vow to honor the results of the election? And you always want to say, like, well, if you don't know what happened, how can you vow in the future? Even a well-intentioned person can't answer that. Yeah. Because you don't know. Because what if there is fuckery? Yeah. Right. What if it's obvious, egregious fuckery? Yeah. And that's the thing that I think people are going to have to be. And by the way, I do think that preparations for fuckery have been underway. And I do think there is structures in place to be able to combat that as much as, you know, they would like to muddle the picture. I do think there are people out there fighting. And I give them the man, the people. that run our elections in this country, the individual citizens, who put themselves on the line to be the administrators of our, you know, democratic republic, like. Yeah. God bless. They are doing the angels work as far as I can see. Yeah. Thank your poll workers. And I think the good news is I saw on Truth Somewhere that if you do see fuckery and you report it, you get a million dollars. What? So. Where's my phone? I wonder if you'll get a million dollars if you report Trump's fuckery. If the money comes from him, he's like, you got me. What else? What else, Brittany? John, if you had to, would you? Yes, I would. I would twice. Now I'm desperately afraid of what's coming next. What is the would you? That's it. If you had to, would you? Oh, yeah. Twice. Yeah, twice. Next question. Wait, was that really the question? That was the question. Oh, God. It's like tautological a little bit. No, hold on. I want to change my answer. Too late. We've moved on. Son of a bitch. John, with everything he has going for him, why doesn't Trump complain that he's never been voted people's sexiest man alive? I think even that. I think even he understands. A bridge too far. I mean, he has a fucking, at this point in his life, he has a front butt. Well, he loves asses. He's got one in the front and one in the back. I don't know. Have you seen that? Back butt though? He loves asses, but he's not paying attention to the front butt. And that's, that has been one of the strangest twists in this election cycle. And it's so interesting in the previous times where he's been kind of the lead campaigner for his own campaigns and everything else. He's always done shtick. There's always been within it, he'll find like a little bit and he'll hit it a couple of times. The enemy of the people or he'll do a little shtick. Like they always say, I watch these things now and I go, it's really all shtick. Yeah. It's not. There's not a lot of air there. There's really, every now and again, it'll be like, even when he's talking about what he thinks are policy wins for his administration, they're shtick because it's, he can't just say, ladies and gentlemen, you know, oil and gas companies are making billions, trillions of dollars. You've got to reelect us to keep this gravy train going for ExxonMobil. So he couches it in. They tell me not to say anything about it, but I do. And the fake news media. It's, it's all now really, he is a Vegas nostalgia act. Yeah. All that's left is like the cult of personality. Like there's nothing on top of it anymore. Right. The accomplishments are nil. And so the shtick has to carry on. Carry the day because he fucking has nothing. None of them do. And the Republicans that are running, what are they running on? We didn't stop him once. Like, we didn't provide any check or balance. Elect us again. Yeah. I mean, that's the hope is like, you still love this guy, right? That's all they're, that's all they're running on. But the shtick is creepy now. Like it's gotten more, forgive me for saying this, but it's more sexual, like constantly. Poor John Husted. Like he doesn't, he didn't ask for this. But let's never forget when it wasn't shtick, it was always wildly offensive. And remember, this is a guy who recovered from grabbing by the pussy, who came down an escalator. And the first thing he said was, I want a travel ban on Muslims and Mexicans are, you know, some of them are okay, but rapists mostly. So he's never tempered those, the vileness of his id. But now he couches it in a kind of comic persona that he no longer carries the straightforward, to use that word that we thought about earlier, vituperative persona. Now it's all couched in, you know, and then he's like, I'm just a harmless old man who likes to stare at asses. Yeah. But still am terrified of killing vulnerable communities throughout this country. And isn't that wonderful? So what else? One week. One week. One more day. Last question, John. For Halloween, do you sit by the door to give out candy or do you just put out a big bowl of candy and trust people to do the right thing? So in, first of all, there is no right thing on Halloween. It is basically, it's a free for all. So if there's, if there's candy in a bowl and you want to dump, it in your pillowcase, God bless, have at it. It is. Those are the rules. There are no rules. There are the rules. The only rules are if someone gives you an apple, you can egg their house. That's really the only, that's the only rule that's going. Unless obviously it's a caramel apple or a candied apple, which by the way, what is the red candy on a candied apple? It is not certainly a flavor known to man. No. Or a substance really. Is it microplastics? It's got to be. Yeah. The shine on it. The shine on that. That's a natural. Is it, is it just melted water bottles with a little bit of a grenadine added to it for coloring? It is delicious. So in New York, so New York was great because the kids, we could go right out on the street and you know, you could do, you could sit on the stoop. That's, we did that one year. We sat on the stoop and just passed out candies. People went buying. That was great. Suburbs are a different deal, man. And it is definitely like now that the kids are older, it's, it's just not a, it's not even really when your kids aren't doing it anymore and nobody really comes to our house because obviously the, the guard turrets and the guns. So that's when you buy the candy just in case and then you eat all the candy yourself. That's right. That's right. Very nice. And I want to remind people our podcast next week. What do we do? We come out normally on Wednesday. I think we're coming out on Thursday because of the election. We want to tape it on Wednesday so that we have a chance to have some sense of, of what happened. So I think the podcast will come out on, on Thursday there. Would that be correct? That's correct. I now lean on every, it's so funny to me during like, during the interviews, like Jillian will just pop in with shit like, you're completely wrong about that. The answer is that, and I'll be like, oh, thank you. Hardly. All right, kids. Well, by the way, how do they stay in touch with us? Twitter, we are Weekly Show Pod, Instagram threads, TikTok, Blue Sky. We are Weekly Show Podcast and you can like, subscribe and comment on our YouTube channel, The Weekly Show. I'm Jon Stewart. Thank you very much. As always, producer Brittany Mamedovic, producer Jillian Spear, video editor and engineer Rob Vitola, audio editor and engineer Nicole Boyce, associate producer Rebecca Rotenberg, our executive producers Crix McShane and Katie Gray. We shall see you next time. Happy election. The Weekly Show with Jon Stewart is a Comedy Central podcast. It's produced by Paramount Audio and Bust Boy Productions. And don't forget to get 40% off Ground News at ground.news.js. Many have been to Deep Ellum. Few have done a deep dive. So if you haven't been deeply moved by a mural, made cheese at the mozzarella company, sampled global fare at Rocket Fizz, dined under a Dallas skyline, lost yourself in a good book at Deep Vellum, or found that the blues make you happy, well, you've barely dipped a toe into a district full of interesting discoveries. Deep Ellum. Go deep. Discover more at deepellumtexas.com. Paramount Podcasts.