Couldn't build more data centers if we wanted to (maybe)

Daily Data: The asset class with the hottest rents, has a bottleneck on new supply

In today’s dispatch:

• there’s one class of real estate where rents are really cooking

• AI as a share of non-residential construction investment

• so they’ll just build more, right?

• maybe the nation’s grid has an even bigger challenger than data centers

---

👉👉👉Reminder to sign up for the Weekly Recap only, if daily emails is too much. Find me on twitter, for more fun. 

---

Why can’t we just build more data centers?

It’s (again) hard to understate how much capital is currently riding on “statistical learning (‘AI’ if you prefer)” right now.¹

Rents are cooking

Consider the case of Data Centers, where rents have increased 20-55% over the last eight months (via CBRE):

Rents are exploding pretty much everywhere, but have nearly doubled in SV.

That rents are growing somewhere, is generally good news to the real estate industry that is otherwise treading water (as rates and pricing settle, and new inventory gets absorbed).

According to Goldman, data centers are currently ~25% of all (private, non-residential) construction spending, and are expected to be closer to 40% this year and the next.

Data centers went from being an afterthought to one of the only games in town.

Point is that AI is as important to real estate as it is to everyone else—and not because they’re using it, but because they’re building it, or trying to.

AI isn’t a tech play, it’s the everything play.

Want to build more, but can’t

That’s the other interesting thing: rents are exploding despite all of the money pouring into building new data centers. You would think that with more supply, prices would come down.

It hasn’t happened yet, though. Why?