What are we getting? I don't want to say 'we' — what are the bulls out there getting wrong about AI? Because you've made no secret that you don't see a 'there there.'
There there, yeah. So I think people are conflating a semiconductor rally with an underlying successful business, which doesn't really exist. And so take their current revenue growth, which is deeply questionable given the fact that they leaked profitability numbers manipulated by Elon Musk. Of course their growth is coming because people cannot measure how much an AI task actually costs. And a couple of months ago, Anthropic started charging their enterprise customers the actual token rates. What this has led to is suddenly businesses going, 'Oh, how much money are we spending?' Uber's — see, I think it was Uber's CEO Andrew McDonald said that they're having trouble justifying the AI spend based on the actual return that one could actually measure. So you've got a thing where you can't measure the costs and you can't measure the return on investment. What do you call that? You call it a thing without an ROI.
Isn't it, though, safe to say — and I think about this, I think that there is going to be — when there's any new technology, there's a lot of money thrown at something, right? A hot new idea, whatever it is. And that ultimately, like we've talked with some experts in the AI world, they're saying all these large language models, not everything's going to exist, because right now it doesn't cost you anything to use it. But at some point there will be a cost and people will start to select what they want to play with. And so is it a case of — I mean, at some point there will be some winners, some losers, and we'll see some things filter their way out, and that there will be some return on investment for some of the larger players. When it comes to AI, do you not see some winners and losers in this, or some winners in this?
I think the winner is Jensen Huang, and the winner is the construction firms who have got prepaid for all the data center construction, maybe Sam Altman and Dot Rio, and they have become billionaires through this. But when it comes to the actual businesses, you can't find anyone who can measure the ROI because you can't do it. So when it comes down to the model companies themselves, they're horrifyingly, horrifyingly unprofitable. I think OpenAI lost like 20-something billion dollars last year. Anthropic is probably not far behind that. We don't have access to their books, of course, but this is based on filings for an IPO. And that's the thing: when we see these ones, I think it's going to be kind of a massacre, because I think that people have this view that these companies are becoming more profitable or even have a path to profitability, and they don't have one.
But, you know, somebody could say — and, Carol, we talk about this all the time — Amazon for years didn't really have profits. Well, that's the thing: now it's a behemoth. And now it's a — you know, then AWS came along, and now it's this juggernaut that, you know, has made a lot of people very wealthy.
But the thing with Amazon is, between 2003 and 2017, all of Amazon's CapEx was around $57 billion, normalized for inflation, in that entire time. And they were cash-flow positive in '09.
The classic dot-com thing as well: things went poorly, but after that there was useful infrastructure. No such story exists here. GPUs — AI GPUs specifically — have very few other use cases. And on top of that, if you're talking about incomplete data centers, it's going to cost just as much money to finish them today as it will in ten years. And similarly, a Blackwell GB200 NVL72 rack, for example, will cost just as much, if not more, in electricity ten years from now. There is no pick-up-and-fix-this story. Perhaps there will be some on-device large language model use, but we are talking an industry that has absorbed over $1 trillion in the last three years. The media attention from everyone, and these continual stories about the magic of AI — but when you say, 'Hey, is it making any money?' they go, 'Oh no, oh, we couldn't possibly. We don't do that yet. But don't worry.' Uber burned about $32 billion, which is less than half of what Anthropic has raised in the last six months.
So I think it's safe to say there's a lot of smart investors in this, smart companies who are in it, right? I mean, pick your big hyperscalers. They obviously want to be around for a long time. So I'm just like, how do you explain their bets on all of this?
So I have this theory called the rock-com bubble, which is the hyperscalers are out of hypergrowth ideas. We haven't had a new iPhone. We haven't had a new cloud computing. We haven't really had a new platform that would turn into the next Google Search or Amazon Web Services. So they thought that large language models were going to be the panacea, that they were going to be creators of API growth, that they were going to be the new infrastructure play — load up the GPUs, sell access to everyone. The problem is, no one checked to make sure any of this was profitable. And we still don't know if GPU compute is actually margin-positive, which is crazy considering the amount of money going into it.
And when we look at the companies that actually are selling it — CoreWeave, for example — they burn billions of dollars, 20-something billion dollars' worth of CapEx this year, about the same amount in debt. And their customers are Nvidia, OpenAI through Google, OpenAI through Microsoft, and Meta, who does not have an AI strategy. It's very strange — like that Twilight Zone episode with the scary child, like we're all trying to pretend that Meta actually has a plan here. They don't have one.
So is Apple kind of smart? Because we've talked with Mark Gurman about this, how they've kind of been relying on Google, right, and Gemini, and not necessarily doing — they're not doing any CapEx.
No, exactly. Well, I think what happened with Apple is they released Apple Intelligence, which may be the most — the largest-scale antitrust radicalization in history. My non-tech friends, I've heard nothing about AI from the moment Apple Intelligence comes out: 'Look, this is the worst thing I've ever used.' And I think that made Apple go, 'Maybe not. Maybe we should take a step back.' But their CapEx is minimal. They're spending about as much in CapEx, ironically, as Amazon was spending on AWS way back when. So I think Apple is cautiously holding back and kind of absorbing the ridiculous headline of 'Apple's behind in AI' — behind in a losing proposition.
We're speaking with Ed Zitron. He's the founder and CEO of EZPR. He's got the Better Offline podcast. He writes the Where's Your Ed At newsletter. Just so everybody's aware: do you have skin in the game, or are you shorting any of these companies? Do you have a financial interest in what you're talking about?
