Is AI Too Big To Fail? The Fed Can't Agree Either

Real Estate Insights. Success Mindset. One Step at a Time.

Issue #63 | August 10, 2026

Three Fed presidents. Same data. Three different answers on whether AI is becoming too big to fail. The debate's about finance. The risk runs through your dirt.

🧠 Market Insights

Is AI Too Big To Fail? The Fed Can't Agree Either

This week, Reuters caught three regional Fed presidents wrestling with the same question, in public, on different days, with different answers.

John Williams runs the New York Fed. He does not see a bubble. His read: investors are still trying to solve, in real time, how big the AI payoff actually turns out to be. That uncertainty is what's driving the swings, not panic.

Jeff Schmid runs the Kansas City Fed. He's watching something more specific than sentiment. He asked whether the contractual chain behind a data center, the commitment to an energy provider, that provider's commitment to the community it serves, is getting too leveraged. Then he asked the follow-up that matters: what happens once something in that chain sparks a problem. He reached for the phrase every banker knows how to hear: too big to fail.

Mary Daly runs the San Francisco Fed. Her read is blunter. Just look at the growth rate and the dollar amount of AI investment right now, she said, and “you could easily say this is very worrisome.”

Three presidents. Same institution. No agreement.

Apollo's chief economist Torsten Slok ran the numbers against the last capex cycle that actually blew up the economy. Hyperscaler spending is on pace to hit roughly 3% of GDP a year by 2027 to 2029, up from 1.4% this year and 0.3% back in 2019. That's still less than half the size the 2005 housing boom reached at its peak, 6.6% of GDP. But the climb, year over year, is running close to twice as fast as housing did at its fastest sprint.

Level says not there yet. Speed says pay attention.

Dirt Dawg Take: All three of them are debating something none of them can actually see yet, how big AI's real payoff turns out to be. What isn't in question is where the physical build has to happen. Every one of those contracts needs a piece of ground under it. If the AI story keeps running, that land is worth more. If it slows down or corrects hard, land with power already on it still has a buyer. That's the position worth holding, regardless of which Fed president turns out to be right.

✨ The Alpha Effect

Commitments Aren't Assets

Schmid's concern is about a chain of commitments: a data center's contract with an energy provider, and that provider's relationship with the community it serves. Those are agreements on paper. A company can revise a deal, delay a project, or walk away from an announcement without much cost.

Land with power already run to it, permits in hand, an interconnection agreement signed, is different. That's not a plan anymore. That's built.

This is the distinction worth paying attention to. Not whether AI turns out to be overhyped. Whether what you're evaluating is still a commitment, or has already become an asset.

Second-order thinking here isn't picking a side in the Fed's argument. It's building a position that doesn't need their argument to resolve in your favor. The acre with power sitting on it holds value if AI blows past every projection. It holds value if half these announcements evaporate next year too.

A signed agreement can unwind. A graded pad with power doesn't.

🚀 One Step at a Time

Action Steps This Week

1. Sort Your Pipeline Into Paper and Poured

Go through every active deal or prospect tied to data center or AI adjacent demand. Which ones have a real interconnection agreement, a completed power study, or a signed entitlement in hand? Which ones are still riding on an announcement? Rank them. The paper ones need a fallback plan before you need one.

1. Add a Community Line to Your Underwriting

Schmid's whole worry is the link in the chain nobody stress tests. Before you go deep on any AI-adjacent site, ask about local sentiment the same way you already ask about power capacity. Has the county held a hearing yet? What happened?

1. Find the Site That Wins Either Way

Pick one parcel in your target market and ask a blunt question. If the tenant behind this deal disappeared tomorrow, does the land still have a second use? Industrial, logistics, another power-hungry tenant? If yes, that's your kind of dirt. If no, you're financing someone else's bet, not building your own position.

Small steps compound. Three Fed presidents can't agree on where this goes. Your dirt doesn't need them to.

Here's to building your success, one step at a time...

from the dirt up.

Kris "Dirt Dawg" McGee 🐾

The Dirt Dawg Den