Intel and Broadcom CEO's Warn AI is Running Out of Capital - AI Data Centers are a Failed Business

Eli the Computer Guy (0:00)

I mean, from CNBC, Intel Broadcom CEOs warn Capital Crunch could become AI's biggest bottleneck. Because obviously, the most valuable technology stack to ever be created, a technology stack with a $30 trillion total addressable market. Obviously, the biggest problems for these technology companies is actually getting their hands on enough money to build out all of the infrastructure they need to build out. So this is a particularly weird article, because the more you read it, the more you realize these AI companies have not actually proven the value of the AI products. So the issue that's going on right now is all of these AI companies have these glorious plans for trillions and trillions and trillions of dollars of build out. But the debt markets, right? The people that essentially loan money, they're getting a little skittish on loaning money. The equity markets, when you buy stocks from companies, those people are also starting to get skittish. The private investors, again, when companies are still private, they're getting skittish.

And weirdly enough, right, we keep hearing about ARR, or annualized revenue, or annual recurring revenue. You know how those numbers are really, really, really big? Like, oh my golly, look at how fast the revenue is growing. You notice how they never talk about profit, like real profit. Like Anthropic comes out with these weird profit-esque numbers, not real profits, like they're... And the problem that they're running into, right, is they're running out of money that they can beg from people, and their products actually are not that profitable, probably not profitable at all. And so they are running out of money to do these absolutely insane build-outs. And yeah, it's... This article just gets so stupid at the end. It's like, when you get to the end, you're just like, what the hell? The chief executives of Intel and Broadcom warned that access to capital could emerge as a major constraint on the AI boom, even as demand for computing infrastructure continues to outstrip supply. And I really want you to think about this for a second, right? We hear about this demand all the time, right? The AI is compute constrained, right? The only thing stopping, you know, more artificial intelligence is literally more hardware resources. But when you start looking at what these AI companies are providing, a tremendous amount of the services that they're providing that are burning tokens, burning cycles on their hardware resources, are literally free, are literally free. It appears, like when you look at Anthropic or OpenAI, it appears something where between 40 to 60 percent of their user base are free users.

On top of that, even when they have the people pay money for the tokens that they consume, it appears even then they're losing money. So if you spend $200 a month on an OpenAI, a ChatGPT subscription or a Cloud subscription, from everything that we can tell, the companies are still losing money on that.

And not only that, as I've talked about before, we were talking about Anthropic a little while ago, you have companies out there, they're spending millions a year on Cloud. One million to over $10 million per year on Cloud. They have these contracts, so they have contracts with volume discounts. And one of the weird things with that is, as soon as they get near the end of that volume discount, whatever the bulk purchase is, they have sales agents calling them immediately, basically threatening them that if they don't sign new contracts, well, they can't be held liable for whatever the prices that they're going to pay. Which again is weird. Look, if you want to fuck over your $20 per month users, again, I don't recommend it, but I get it. Fuck them. They're $20 a month users. Like, can you imagine strong arming your $10 million a month user? That seems a little weird, right? And so that's one of the bizarre things with this, is we hear about how much demand there is for AI, yet most people, the vast majority of people, don't want to pay for AI. And even the people that want to pay for AI are not even willing to pay the amount that would be required to cover the inference costs much less anything else. And so again, we go back to this whole, I'm sorry, I don't really understand the demand argument here, with artificial intelligence. Broadcom CEO Hock Tan said, financing could become a bigger obstacle to AI expansion than shortages of semiconductors, memory chips, or electricity, as the industry races to build infrastructure needed to power increasingly sophisticated models. And again, we go back to this whole thing of do we need more increasingly sophisticated models? As I've argued in 2026, I think the orchestration layer, the control plane, is the most valuable part of the AI stack. Do we really need better and better and better models in 2026? Is it worth investing? Here's the thing. If they had to invest purely off of revenue and profit, would investing in new models actually make sense right now, right? Since they've been getting so much investor money in, right? They can just burn investor money, as I've talked about before, right?