The $1.7 Trillion Data Center Bet, and Why Apple is Getting Into the Subscription Game

The Prof G Pod with Scott Galloway

Scott Galloway examines the massive $1.7 trillion debt bubble surrounding AI data center construction and the potential for a market correct

Key takeaways

• The AI buildout involves an estimated $1.7 trillion in hidden debt, much of it pushed into off-book shell companies.
• Bloomberg predicts that up to half of the U.S. data centers planned for 2026 could face delays or cancellations.

Main topics

• AI data center debt risks
• Apple's subscription strategy

Notable quotes

The build-out runs on many borrowed money right now.

Conclusion

The episode highlights how strategic shifts toward recurring revenue a

Transcript preview

Speaker 1 (0:05) Welcome to Office Hours with PropG. This is the part of the show where we answer questions about business, big tech, entrepreneurship, and whatever else is on your mind. If you'd like to submit a question for next time, you can send a voice recording to officehoursproptgmedia.com. Again, that's officehoursproptgmedia.com. Or post your question on the Scott Galloway subreddit, and we just might feature in our next episode. Question number one comes from a listener who emailed us. Hi, Scott. I'm an ex-UK Army engineer officer working on grid upgrades in the Scottish Highlands. Your No Mercy, No Malice episode, 1999.ai, has led me to believe that the push on data center construction is ill-advised and will end up with data center developers holding the crying baby with incomplete sites and sunk costs. There's also the question of realistic capacity in the grid and the demand from new building housing and other sectors. Is the model of speculative build of data centers for hyperscalers destined to fall down? Thanks as always for speaking out, swearing, and drinking. Well, I won't stop that. Your ancestral home is proud to call you a jock. Huh. Well, you oatmeal savage. That's what my father used to call me when he was joking. Some data. Most announced data centers were never real. To build a data center, developer applies to plug into the power grid, and operators get five to ten times more applications than real buildings because they file a whole pile. to hedge. Most have no financing, no power deal, and no equipment order. And in some, most are most planned data centers or press releases. Bloomberg expects a third to half of all U.S. data centers planned for 2026 to be delayed or canceled. Why people seem to, the latest scare around AI is the debt. Wall Street's mood has flipped fast. In Bank of America's investor survey, 45%, now called the AI bubble. the biggest risk, up from 28 % a month earlier. The build-out runs on borrowed money right now. One estimate puts the giant's hidden debt at $1.7 trillion. That's up eightfold in four years. The scary part is what's not on plain sight here, and that is $800-plus billion is pushed off the books into shell companies, so it's hard to see who's going to eat the losses. Plus, this debt is backed by chips that can lose value overnight if a better one emerges. A lot of people are equating this to 1999, and there's real parallels here. Ordinary people, in my view, will be the ones left holding the bat. I think there's going to be an AI slash data center bailout from Trump, who's bet the entire economy on the valuations of AI to continue to be out over their skis. In many states, utilities are guaranteed a profit for building power plants, so if the demand never shows, it lands on everyone's power bill. The financing is circular. NVIDIA sells the chips and lends its customers the money to buy them. And so the same dollars keep flowing in between a handful of companies, making demand look bigger, artificially bigger than it is. In sum, the end here or the slowdown isn't likely one dramatic bang. It's kind of a slow rolling thud similar to 2002 where the weakest, most borrowed players fall first while big tech survives. So this listener, you're right about the crying baby and who'll end up holding sort of the over leveraged developers and the rate payers, not the hyperscalers. But my understanding is you're working on grid upgrades in the Scottish Highlands. I think it's interesting to be in the center here in Inspector. I would argue, as it relates to your job, I still think the demand is going to be pretty substantial and that your skills are going to be in pretty high demand for a while. In other words, the technology, and I do think the data centers, will survive. I just don't know what the valuations will be or who pays off the debt. to finance these things. But as it relates to what you're doing, I think you're in a pretty good spot right now. One, working in the Scottish Highlands, I mean, I have an ability to romanticize Scotland because I think it's the most beautiful place in the world. So to be there, and I don't know if you're working outdoors or probably not. I don't know, to have any skills around adjacent to data centers right now is a pretty good wrap, probably for a few years. I just don't think these things are going to, I don't think we're going to stop. building data centers. I just wonder if the people financing this type of growth are going to scale back or their debt is going to come under pressure. But brother, to be working in the highlands and have skills around what is at the center of AI, and that is the data center build-out, I still think there's going to be enough demand to justify the centers that do get approved, if you will. Anyways, very much appreciate the question and congrats on your geography and your profession. Speaker 1 (5:01) Question number two comes from Jacob. Hi, Scott. What do you think about Apple's move to a subscription-based model? To me, the subscription model not only means more reliable revenue for the company, but it is also stickier. When consumers get on the plan, they'll likely be content just to keep receiving new iPhones, regardless of the fact that the only substantial upgrade year over year is a better camera. Apple knows they are done innovating in that space and they want to get lock-in customers now. I'm curious about your thoughts. Yeah, I think we're in violent agreement on this one. On July 28th, Apple launched an Apple upgrade in the U.S. The program will let customers pay a monthly fee for an Apple product, then upgrade to the newest model, return the device, or buy it outright when the lease ends. Leases start at $18 a month for an iPhone. This is likely a reaction to a slowing upgrade cycle. The average global smartphone replacement cycle is expected to stretch to four years in 2026, up from three and a half years in 2025. In sum, the improvements have become incremental. And I think people have decided they don't need the latest and greatest right when it comes out. Apple's subscription business is working. Apple's services business, which