Etienne Note: Zitron didn’t mention one of the major aspects of this story. If Microsoft (and other companies) hasn’t installed over 2/3rd of their purchased infrastructure without any impact to their existing customer usage, are they really expecting demand to triple when all of the heavy hitter Fortune 500 customers are already using AI?
Transcript by The Art of Liberty Foundation
Ed Zitron
like half of NVIDIA's revenue could literally be sitting in warehouses. I thought it was tens of billions of GPUs, but based on this Microsoft story, could be hundreds of billions. If I'm right here, this is going to be one of the greatest financial shenanigans for tech. This is going to be an absolute scandal in a way that will humiliate some of the largest companies in the world. NVIDIA's claiming they're going to double sales next year. If that's the case, OK, how many of those are installed in data centers? Because if it's half or a quarter or a tenth, or it's going to take three or four or five years to install them, How do we actually value this boom? Is this a boom?
Isaac Pound
Hi, I'm Isaac. And on the tech report with me today is writer of Where's Your Ed At? I'm the host of the Better Offline podcast, Ed Zitron. Welcome back. Thanks for having me. So just before we started, we completely did away with the original question list that I had, because you sent through some, some interesting stuff from Bloomberg, who's reported that Microsoft only has about two gigawatts of actual AI compute online out of an expected 12 gigawatts that's supposed to have by now. I mean, just, yeah, just tell me, tell us about that.
Ed Zitron
So Bloomberg said two gigawatts of IT load of just AI chips with Microsoft. And Microsoft has been saying they've been adding one gigawatt capacity every single quarter for like three quarters, I think it is. I think they said they doubled their capacity in one of them. And it's funny, a few months ago, Ayesha Down and I at The Guardian reported a very similar story and said that Microsoft only had a few million chips. It was less than they expected. And when we went to Microsoft, they said, oh, this is non-indictive. You don't know what you saw. So the source I talked to and that we talked to I actually had this number. I had 1.993 gigawatts and a precise breakdown of the chips. And the amount that they're worth is about $50 billion. Microsoft has spent over $265 billion in CapEx. And if we assume that CapEx, if like 50% of CapEx is GPUs, it's actually impossible to tell these companies don't disclose. That means that Microsoft has like $82.5 billion of GPUs not installed at all. Now, this is significant for Microsoft. I'm going to do a newsletter on this next week. Significant for Microsoft, but also for the rest of the world, and also begs a few really worrying questions. Microsoft was the first company, I believe, to build an AI supercomputer. By which I mean, in 2020, they built one out of A100 chips from NVIDIA. There were other AI supercomputers, but it was the first big one out of using the Mellanox-powered, Mellanox was a networking startup that NVIDIA bought to network together GPUs, the first one of those in 2020 for OpenAI. Now, they have the most experience, they have the most logistical experience, they have the most money to do this. Amazon technically has done more cloud for longer, but Microsoft is way more in AI. And Microsoft has what looks like less than half of their purchased GPUs installed. This means, I believe personally, that this is an across the industry phenomenon. And if that's the case, we likely have hundreds of billions of dollars of NVIDIA GPUs in warehouses waiting to be installed, which begs the question, why are people still buying them? And how is Jensen Huang doing it? Now, big boy Jensen Huang, CEO of NVIDIA, might just be fandoming them. He might just be saying, you pig, if you don't buy these today, I won't sell you the next ones. But this is fundamentally insane. This means that GPU capacity is coming on extremely slowly. And I believe, personally, that there is an industry-wide attempt to lie about available capacity. I think Oracle claimed they brought on several hundred gigawatts, sorry, megawatts even, worth of capacity on last quarter. They claim there are six out of eight buildings in Stargate Abilene based on discussions with the source less than a month ago. That's not true. There are less than, I don't even think the fourth was brought online. And so everyone is talking about capacity. We brought this much capacity on them. That's this much capacity. I don't think they're being honest. I think what's actually happening is there is capacity being brought online as in CPUs and infrastructure related to GPUs, but not the GPUs themselves because of power. of data center construction times. But here's the thing. Is NVIDIA running the largest Kickstarter of all time? Is NVIDIA just selling GPUs years in advance? How many GPUs have come online? Because if it turns out that everyone is like this, everyone is just hoarding GPUs, that is actually a That should be an SEC investigation, because put it like this, I don't think it's illegal, but it's something that I think shareholders need to know, because there's a big difference between NVIDIA is selling a bunch of GPUs because of demand for data centers that are getting built, and NVIDIA is selling a bunch of GPUs because everyone's worried that if they don't buy them now, they won't be available in the future. And I must be clear, Microsoft is so experienced at doing this. If Microsoft's got less than half of their GPUs installed, do you think Meta's much better? Do you think Amazon, Google? I don't think so. Google's apparently building 4.5 gigawatts worth of capacity for Anthropx TPUs, so Google's AI chips, to which I say, no, they're not. Like, I actually question how many data centers have been finished here. I brought this up a few months ago, but these numbers really bring it home. Only 2 gigawatts of capacity. That's crazy. And I must be clear, when I'm talking about the capacity, having seen the numbers I have, a lot of that is H100 GPUs and H200 GPUs. It's older GPUs. And this includes data center capacity, I think, from either N scale or at Nebius. It doesn't include Core Wave. Long and short of it is, is NVIDIA just pre-selling GPUs years in advance? Because if that's the case, there isn't AI demand at the scale. The reason that we have capacity constraints is stuff is not coming online at all. Like if Microsoft, I mean, this story, this two gigawatt one, came out a couple weeks ago. I must be clear, the numbers I saw were from months ago, and they were 1.993 gigawatts. Now this was my own mathematics based on TDP, so the power that the GPUs pull, pretty close though. But I think it begs the question of what exactly is going on with the AI data center build out? Because if it's years delayed, If it's truly years delayed, that means that any of the gigawatt figures we've seen are nonsense. Because they claim they brought 12 gigawatts online, only 2 gigawatts is AI data center capacity, like actual chips, so the... I'm guessing it must be CPUs, but... Okay, are we meant to be impressed by a bunch of CPUs? What is it that's actually going on here? We need clarity, and we're not going to get it because we don't have a functional regulator. We don't have functional financial regulators, and we certainly don't have journalists who are going to push much further than this. And it's frustrating because this here is a scandal.
Isaac Pound
And surely it's going to only get worse, because the longer these GPUs are just left sitting around collecting dust, the less relevant they become. And then it's going to reach a point where everyone's like, oh, we've got this amount of gigawatt, we've bought all of these GPUs. Oh, actually, yeah, 50% of that was actually, we never installed them, we never used them.
Ed Zitron
But that's the thing as well. It's, okay, so you've bought tens of billions of dollars of Blackwell. By the time you install all the Blackwell, we'll be on Feynman, which is the thing after Vera Rubin. By the time you install Vera Rubin, it'll be LMFIO Seal. I don't know what they're going to call the next ones, but they're going to call them something. Also, I'm repeating myself, how much capacity is actually online? Because I've seen, oh, 17 gigawatts of capacity has come online. I've seen 12. I've seen 10. But if that 10 gigawatts is only 1 gigawatt, just to be clear, this 2 gigawatt number is all time for Microsoft. This isn't just for online. This is all they've got. If the total amount of AI capacity that's come online to date is only three or four gigawatts, that's crazy. And this is why I think all of these capacity stories are so ridiculous. Because if it turns out most of it isn't AI chips, then those numbers are functionally useless.
Isaac Pound
So yeah, just take us through, what is the significance of the fact that instead of it being 12 gigawatts, it's just 2 gigawatts?
Ed Zitron
Well, it means that Microsoft has been boasting about building gigawatts of data center capacity, with everyone thinking, oh, they mean they're building AI data center capacity, and they're not. means the AI data center capacity just isn't coming online. It means that they have been sinking hundreds of billions of dollars into something, mostly to warehouse it. Now the argument would be, okay, we bought the GPUs, the construction's in process, it'll come online at some point, but when? And if Microsoft, the apex predator of AI infrastructure, cannot do this, you think Meta isn't doing the same thing? Meta, what, 130-something billion, 160 billion they're gonna spend this year? Little under 100, I think, last year. Amazon was, I forget, maybe 200 last year. Are those chips coming online? How many chips? How much of that capacity is actually AI infrastructure? Because if it's not much, then we have seen the trillion dollars of CapEx spent on bringing online, what, $100 billion, $200 billion worth of chips, as NVIDIA has sold $535 billion worth of them. And I think that when you tell people this story, I've tried to tell a few reporters this, they go, eh, you know. Eh, it's not a big deal. It's a huge deal on actually an insurance level. If you've got hundreds of billions of dollars or tens of billions of dollars of chips sitting in a warehouse somewhere in Taiwan and in America somewhere, you need to insure that. And I don't know, I hope the warehouse is good. I hope nothing happens to the cooling in there. I hope that nothing's wrong with the chips when you turn them on. Also, why buy so many in advance? Like, that is the real question here. Why have you pre-ordered so many? And sure, you could say, well, I need to make sure I have the capacity in time. But if you're taking years to build the thing, well, you could have waited. You could have waited a while, unless, of course, Jensen Huang is just fin-doming them, and he's saying, if you don't buy this pig, I won't sell you GPUs in the future. And I think that's been very effective. But if it turns out, like, half of NVIDIA's revenue could literally be sitting in warehouses. I thought it was tens of billions of GPUs, but based on this Microsoft story, it could be hundreds of billions. It could be the vast majority of its 2025 revenue. That's a crazy number. And NVIDIA and Microsoft and all of these companies have been working really hard to obfuscate the actual numbers. Let's think of it like this. Microsoft, every single quarter, I think it's three straight, said one gigawatt capacity added, one gigawatt capacity added. It's very clear that within that one gigawatt, there's what? 50 megawatts of capacity? 100 megawatts of capacity? It's horrifyingly misleading. And it also misleads shareholders into believing that this capital expenditure fiasco is leading to actual revenue growth or will lead to anything in the next few years. And the question is as well is, why do you buy Blackwell when Vera Rubin's coming along? Why do you buy Vera Rubin when Feynman's coming along? And if the answer is, NVIDIA won't sell it to us, well, that sounds like a predestiner's dilemma situation to me. That sounds like something that's entirely created by everyone buying the GPUs. If everyone went, all right, I've got plenty. I'm going to build some in a year. I'm going to need them then. And if everyone just kind of backed off buying these in advance, everything kind of ends for Nvidia. I think that that kind of happens anyway, because the amount of debt they're going to need at a time when interest rates are spiking is, it's going to be kind of monstrous.
Isaac Pound
That's one of the, I mean, this all kind of raises one of the big questions that I've had in the whole data center build out, which is that even, even if you're super optimistic and you get everything built in two years, which is, Yeah. Very optimistic. Not going to happen already. It's already past two years. Then you're buying GPUs from right now to put in data centers that are going to be built in two generations time. Yeah. Two GPU generations time, which to me doesn't make any sense. You don't want to be buying today's GPUs at today's prices when you could be buying them in two years at two years prices. And you'll still be plugging them in at the same sort of time anyway. And then it also begs the question of why spend even more on future CapEx if you've got all of this backlog waiting to be plugged in that you're, yeah, not even plugging in apparently.
Ed Zitron
That is actually on a very reptilian investor level, on an ugly one. That's a fair question. Those could have gone into share buybacks. I hate share buybacks. I think they're evil. I don't think they should exist. But if you're being an investor and you're saying, where does the company pay back its shareholders? Well, why are you buying years worth of GPUs in advance when you could be paying back shareholders? This will be the argument that is eventually made these companies. That trillion dollars you could have put into share buybacks, why didn't you do that? Why don't you put it into R&D? Why don't you put it into something useful? And it also begs the question of how Broadcom is growing. How AMD is growing? How many chips are actually installed? Because the argument NVIDIA could make is that the power system that you will need, like the actual cooling you will need, will be specific to a generation. That's only strictly true with Blackwell and the first generation of Vera Rubin. They both use the same housing, but then the next generation of Vera Rubin will acquire its own. So perhaps it needs that, but I don't know. If NVIDIA knows how that power is going to work, why do you need to buy Blackwell today? I think it's just we need a real investment. I call upon all tech and business journalism. It is time to find out how many GPUs are actually installed, because these gigawatt numbers, these megawatt numbers are nonsense. They are nonsensical. It's very clear that Microsoft has not brought a gigawatt of capacity online every quarter. And what's funny is in the Guardian story, I may have said this already, but I need to repeat it. They said, oh, we've misread the numbers. Oh, it's totally wrong. Don't think so. I actually think we read them completely right, which begs the question as to why Microsoft wants to muddy the waters and how much everyone else is doing so too.
Isaac Pound
Obviously, this is incredibly speculative, but if we take what is happening here with Microsoft and we do kind of apply it to the rest of the industry, how much very rough figures, obviously, we've just kind of come across this, how much do you think of the capex of the AI bubble? The whole kind of thing is, built on this assumption that everybody else has been plugging in more of the GPUs than they say they are?
Ed Zitron
I think all of it. I think that the massive sales that NVIDIA is making are based on the idea that there is massive demand for the GPUs. The massive demand for the GPUs appears to be pre-orders. I had an accountant friend who asked, do those GPUs even exist? I hope they do, because if they don't, that's super bad, legally speaking. I'm going to assume they do. I wonder if they even left Taiwan. I wonder if there aren't, I, this thing, $535 billion is about how many, it's, it's, that's the entire revenue of the data center segment, which is pretty much all GPUs, but it's got some networking in there too, since 2022 fiscal year. So. The question, I think within that low end, 100 billion in warehouses, high end, 350. And I know that this is entirely speculative, but if Microsoft has only brought two gigawatts on in all time, They have very likely only brought on a small share of the Blackwell GPUs they've bought. If they are in that position, I would be shocked if Meta has brought on more than 10,000, 20,000 worth. I also question how much capacity Nebius, Iron, CoreWeave especially has brought online. Because CoreWeave claims they've brought on like 800 megawatts in the last year. Have they? their revenue doesn't even necessarily suggest that they have. So what's going on there? And if this is an industry that is basically just a pre-order campaign, investors deserve to know. Because There's a vast difference between NVIDIA is selling GPUs because everyone wants to build data centers and they're getting built and they're selling the capacity. And NVIDIA is selling GPUs to people that may or may not install them one day. Because one is a reflection of demand for the data center capacity, which I think is overstated, but nevertheless, if the data centers are getting built and ready to open AI and Anthropic, that's still demand and revenue. If the answer is it's taking years more than we expect. That means that the hangover from this is gonna be really brutal because it's just, it will either all come online at once or come online in drips and drabs and eventually just stop, or it never comes online and they have to write off all of them. And no one has a good answer. The companies won't give one. I can't find many journalists. Aisha from The Guardian, respect to her. She's one of the ones who took this seriously. And I think what we need to do is we, as a journalism industry, is we need to actually investigate these companies and how much of the capacity they have brought online is actual AI chips. Because if it's a very small fraction, the people deserve to know.
Isaac Pound
And then we kind of touched on it a bit earlier, but what happens when, sure, maybe you can skip a generation of GPUs for your training. You skip a year or two, push it to three, whatever. What happens when you get to that point, you're like, I can't put off the upgrade any longer. and there is this glut of GPUs that were never plugged in.
Ed Zitron
That's the thing. They're saying, oh, we have way more demand than our capacity can handle. Again, big difference between the reason that we have more demand than it can handle is because the stuff isn't turning on, or, wow, we have so much capacity and we're selling all of it. Because it can't really be both. It has to be one. And if Microsoft is If Microsoft is barely bringing it online, that means that at some point they're going to have a glut, or they're going to just write them off. But I think it's more likely that they just slow roll this until they eventually, because until the GPU, I believe, accountants will roast me if I'm wrong on this one, I believe the way it works is that if the GPUs are not brought online, the depreciation doesn't start. Microsoft, on an accounting level, would eventually... Because if they're not brought into service, the depreciation schedule just doesn't begin, because their useful life hasn't begun yet. That being said, in the event that anything happens to OpenAI and its ability to pay, Microsoft will have to write them off, legally speaking. But that assumes that Amy Hood is actually going to... I think Amy Hood, who is a great CFO, is going to stretch this as much as humanely possible. And the thing is, this shouldn't be legal. We should, hey, here's an AI regulation for you. Hyperscalers have to say how many chips they have and how many of them are online. Real simple. And people will say, well, they shouldn't have to do that. That's unreasonable. Absolutely not. When you're spending a trillion plus dollars, you should have to show me your insides. I should know everything about you. I should know what you have for dinner. I should know at the very least how many GPUs you bought and how many of them are online. Because right now, I mean, but we also don't, another thing is we have no idea where this CapEx is even going at this point. We don't know. I mean, we know it's going into GPUs, but is it 50%? Is it 40%? Is it 70%? We don't know. They won't tell us. And if it's just, we're buying a bunch of GPUs to sit in stock, that is actually an asset issue. That is a fair valuation issue. How do you value Microsoft stock if you don't know whether those GPUs are going to ever be installed, or whether they're being installed in dribs and drabs? It makes valuing these companies much, much more difficult, and it's unsafe for investors. Because if at some point, if at one moment, these companies have to go, all right, we're not going to use these things, talking $20, $30, $40 billion write-offs, and they'll spread them out, they'll find funny ways to get around it. whole situation is very ugly sorry if this is all over the place it's just this this whole thing really bothers me because this feels like the stripper with five houses from the big shore This feels like the thing of like, hey, we're selling a bunch of GPUs, but we're not actually using them yet. And we're buying them because Jensen Huang says he won't sell them to us unless we buy them now. It's kind of like those people like, oh, I can't have a kid, so I'm going to immediately rush to have one, and they have one immediately. Well, maybe not identical. I don't know. It's just, it's frustrating because If I'm right, and there's always an if, we're speculating. If I'm right here, this is going to be one of the greatest financial shenanigans. This is going to be an absolute scandal in a way that will humiliate some of the largest companies in the world. Elon Musk claims he's brought a gigawatt of capacity online. How many of those are GPUs, Elon? And he won't say or be like, they're all GPUs. It's 100% GPUs, which will just be a classic Elon Musk lie. I just feel like it's insane we never had this clarity, and we deserve it now.
Isaac Pound
It feels like a very tangible sort of representation of the fact... that a lot of what has been going on is hype. That everybody's been saying, yeah, it's happening, it's happening, it's happening, look at how much is being sold, it's happening, you need to buy in, get in, get in, buy, buy, buy. And then being able to then see the figures and say, okay, you were telling me to buy, but actually now I've got no use for these, and actually you've got no use for these, and probably this other guy doesn't have any use for these either.
Ed Zitron
Why have we bought all of these GPUs again? It's also just a disconnection from value fundamentals. It's like, are we valuing NVIDIA based on sales of GPUs because there's real demand? Or are we valuing NVIDIA based on the idea that one day there will be demand? I guess NVIDIA hasn't broken any laws. NVIDIA is sold to what's the quote from Margin Call? We are selling to willing buyers. And the problem is as well is, so okay, let's say $100 billion worth of GPUs, maybe 150 are sitting in warehouses. The cost of RAM has spiked. The cost of CPUs has spiked. The cost of everything related to tech has spiked so that we can warehouse a bunch of silicon. That is insane. That is morally wrong. I'd put aside the capitalism of it all. It is just morally wrong to increase the cost of everything so that you can have things in a box waiting. NVIDIA doesn't need to sell these in advance. But because NVIDIA has them by the gonads, it can. It also... It's actually kind of radiates outward, though. How's AMD doing? How's Broadcom, who's building OpenAI's chip, TPUs, and Samsung, who might be building one for Anthropic? Because if a data center is taking two or three or four or five years to build, then should we be as excited about Broadcom? Does this mean that one day NVIDIA is just going to hit a wall? Because it means that NVIDIA is claiming they're going to double sales next year. If that's the case, OK, how many of those are installed in data centers? Because if it's half, or a quarter, or a tenth, or it's going to take three, or four, or five years to install them, how do we actually value this boom? Is this a boom? And indeed, what is going on long term here? These supply chain agreements with Micron and Samsung and SK Hynix, they've been signed under the auspices that NVIDIA will sell eternal amounts of GPUs, when it seems like NVIDIA is actually kind of the weakest link here. Because to your point, at any point someone could go... you know what just realize something i'm buying gpus from you jensen but i've got 50 100 billion of these bastards sitting in a in a lorry back round back i don't need these right now i'll get back to you when fineman comes out whether that happens when that happens is an open question but this is all very strange and i don't know it's the hype cycle like you said it's the hype cycle which fails to actually evaluate the value of these companies and values them based on what might happen versus what's actually happening today.
Isaac Pound
And also just to, I guess, round things off, maybe the thing that I think about is if you're building a data center, sure, you need to know kind of the architecture of building it around. But if you're like, OK, it's got a three year lead time on this one based on these other ones that we're building. Maybe we buy the GPUs when it's finished construction, when you've got somewhere to put them.
Ed Zitron
Yeah, or three months in advance. Yeah, yeah, yeah. Like, not however many years. Also, if Jensen Huang is saying, I will not sell them to you unless you buy them immediately, that's a material concern for investors. And I think that when you, when we look, here's the thing, who knows how this ends, but when we look back on this era, there could be a decent chance that Nvidia's incredible run was mostly pre-orders. And if that's the case, that's... very strange and also very bad because it's left every major tech firm with some sort of horrible, pestilent pile of GPUs that are useful for one thing, and then a bunch of debt other than Microsoft. Microsoft's actually been quite canny about avoiding that debt.
Isaac Pound
Well, Ed Zitron, thanks for taking the time.
Ed Zitron
Thanks for having me.
Isaac Pound
If you enjoyed 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.
Become a Liberty Member for $5 a Month or $50 a Year and Get Perks
We moved off Substack since they were censoring us. You can read about it and see the evidence HERE.
We just spent $25,000 to build our own infrastructure so we can’t be silenced again.
We are widely exposing the MOST IMPORTANT INFORMATION IN THE WORLD: The illegitimacy and criminality of “government,” monopoly media/academia and the algorithmic censorship of the DARPA Internet from a principled voluntaryist perspective of REAL Freedom, free markets, private property and peace.
Monopoly “Government” has been “The Biggest Scam in History” and has produced The Greatest Theft in Human History! The theft of $1.40 MILLION from the average worker through inflation, confiscatory taxation, and Social Security underpayment. Much more for higher earners! We have the only solution that gets the planet out of voting harder in easily rigged elections: The “One Way Revolution” of widely exposing the illegitimacy and criminality of organized crime “government!”
We are rebranding The Daily News from the Art of Liberty Foundation to Voluntaryist.News where we will continue to syndicate the best of the authentic, developing alternative media while keeping that service free for all as we migrate it from Substack.
Our weekly newsletter, Five Meme Friday, will remain free and continue to syndicate hard-hitting memes, the best of the alternative media, documentaries, censored videos, liberty events, truth music from the movement’s hottest artists, while covering the libertarian intentional communities and free private cities.
We are making Important News from the Art of Liberty Foundation, our investigative journalism and original writings “Paid Only,” and asking our readers to chip in just $5 per month to help cover our rising costs, which gets you access to the ePub and PDF copies of ALL of our books and the Liberator dropboxes!
Become a $50 Yearly member and get a free paperback copy of ANY of my books (or a Liberator), including our NEW book: The Greatest Theft in Human History – How the Banks and “Government” Will Steal at Least $1.40 MILLION from the Typical American Worker… “Legally.” It is the ONLY way to get a physical copy until October.
Become a Supporter ($125-$499), Sustaining Member ($500-$999), Patron ($1000 - $4999), Benefactor ($5000 - $24,999), or President’s Club member ($25,000+) and get perks ranging from signed, high-resolution hard copies, cool stickers, t-shirts, and even a one-day Art of Liberty conference + Turtle Party/Concert in your town with a VIP table for your friends and family!
Sign up here: ArtOfLiberty.org/Become-Liberty-Member
