AI CapEx Boom: Tech Giants’ Spending, Depreciation, and Market Impact

In their latest episode of the VALUE: After Hours Podcast, Tobias Carlisle, Jake Taylor, and John Rotonti Jr discussed AI CapEx Boom: Tech Giants’ Spending, Depreciation, and Market Impact:

John: Yeah. That’s definitely not getting enough coverage. I think the other story is the– I hate to call it this, but the AI trade has reversed dramatically and quickly ever since January 27th, I think.

Tobias: Is that DeepSeek?

John: Yeah, DeepSeek on January 27th. And now, the narrative is that hyperscale CapEx is going to– either the rate of growth is going to decline. So, it’ll still be growing, but it will grow slower. Or, absolute levels of CapEx will just decline from really, really, really high levels. I think they’re expecting $320 billion or $330 billion in CapEx just this year from the Mag 7 or the Mag-10, whatever it is.

Tobias: It’s a big number.

John: It’s a big number. But here’s what’s interesting, and this only take a second. Hyperscale mega CapEx almost never goes down from year to year. It very rarely goes down. Microsoft did $800 million in CapEx in 2005. It’s doing $80 billion [Jake laughs] this year. That’s a 100x. So, in 20 years, 2005 to 2025, a 100x. Google $800 million in CapEx in 2025. It’s doing $75 billion this year. So, the exact same story. Roughly 100x in CapEx increase in 20 years. One more for you. Amazon CapEx in 2005, $204 million. $204 million. It’s going to do a $104–

Jake: $20 trillion.

John: -$104 billion this year. That’s a 500x increase in CapEx from Amazon.

Jake: I thought these were cap-lite businesses. What are you talking about?

John: Exactly. Yeah. That was, I think always misunderstood.

Jake: Well, wait till you run all that through that depreciation through the income statement in a couple years and [Tobias laughs] what do earnings look like?

John: So, that’s really interesting. So, my first point, is these numbers don’t go down often. And if they do go down, they don’t go down a whole lot. Historically, at least, they don’t go down a whole lot. And over a 20-year period at Microsoft and Google they’re up 100x and at Amazon it’s up 500x on the depreciation.

So, these companies are depreciating these assets at five years, roughly. Some, four. Some, six. But they’re depreciating these servers at five years. So, like you said, it’s going to hit the P&L in a few years. It’s going to hit hard I think. There could be offsets to some degree. So, if revenue growth accelerate—

Jake: Write it all down in one year and then–

John: Well, you could do that. [Jake laughs] Yeah, right. If revenue growth accelerates, it actually has to accelerate. That’s one offset. So, revenue growth grows faster than the depreciation line on the income statement.

The other potential offset is, if these companies, and this is what I think is going to happen to a large degree, maybe not offset all of it, is if these companies find internal efficiencies from all of this AI. Google mentioned AI is writing a quarter of all code across the company right now. I think the efficiencies that they could potentially find from their own AI, it’s almost unimaginable at this point. It’s just as new. This is new. And so, I do think in the next few years they’re going to surprise us with efficiencies. Labor cut.

Jake: OpEx down basically from-

John: Yeah. Depreciation up–

Jake: -substituting man with machine.

John: Exactly. Exactly. So, depreciation up big. I think revenue growth is stable-ish, low double digits at some of these companies. Not Apple, but some of the others. So, if revenue growth can be stable and depreciation runs its way through, it’s going to be a big hit. But if they can offset it with efficiencies, then maybe the margin hit, and the earnings hit and the hit to ROIC too won’t be as big as expected. But that’s just one scenario. The other scenario is what Jake said, which is that it just flushes everything out and earnings growth dramatically slows for a period of time.

Tobias: It did look like Microsoft was pulling back. I saw some tweets– [crosstalk]

Jake: Canceled some leases or something. Is that what the– I saw a tweet about that, I think.

Tobias: [crosstalk]

John: It was a lot of verbiage. I think they pulled back on some leases, but then they reiterated $80 billion for the year and they said they’re just rearranging stuff around the world, because they’ve got, as you know, data centers everywhere. It’s just a little confusing from Satya, I think. And then, he put the IR team on it, and they did a conference in Australia and then they put out a press release saying that it’s still going to be $80 billion for the year. So, it was a little confusing.

Jake: Is that what investors want? Are they cheering on they want Satya to spend $80 billion on CapEx?

John: It’s really confusing. If he spends all this 80 billion, then it props up the AI trade elsewhere, all those electrical component companies, the power companies, the servers, the chips, Nvidia-

Jake: All the revenue-

John: -prop up that. Yeah.

Jake: -it comes from few customers, right?

John: Exactly. So, it props up that supply chain into the data center which the market wants. But if you want to see higher free cash flows at the mega cap tech, then you want less CapEx, that’s the other thing. If these companies cut back on CapEx, then they’re going to be printing more free cash flow. It’s nuanced and it’s confusing. But I don’t think CapEx is going down dramatically over time.

You can find out more about the VALUE: After Hours Podcast here – VALUE: After Hours Podcast. You can also listen to the podcast on your favorite podcast platforms here:

Apple Podcasts