Hey Income Futurist,

Most of the AI buildout talk centers on demand. I don’t worry about demand. Every bit of new capacity gets used. The bottleneck that gets less attention is money, and whether there’s enough capital to keep paying a bill that keeps climbing.

Bloomberg put it plainly: “Virtually every major category of costs associated with building a data center is ballooning… The price of short-term memory used in data centers is more than five times what it was just a year ago… Demand for equipment that generates and carries electricity is so high because of the building frenzy that suppliers have backlogs stretching beyond 2030.”

Only $492 billion of the $2.3 trillion data center pipeline is actually under construction. The other $1.8 trillion is planned or precommitted. Those projects haven’t locked in their costs yet, so they’ll pay today’s prices or higher.

The chart from Cushman & Wakefield shows where the inflation lives. Construction materials are up about 7% since the end of 2021. Over the same stretch, switchgear’s up roughly 60%, and transformers, cooling equipment and generators are each up around 30%. Concrete and steel lag way behind. The real cost pressure sits in the gear that gets power into the building.

Sage Road Research put numbers on it. The average cost to build a data center has gone from about $7 million per megawatt before the pandemic to $11.3 million this year. AI-optimized facilities are expected to run $15 million to $20 million per megawatt. Chip spending averaged 0.47% of world GDP for two decades. This year it’s projected at 1.25%, and next year it could top 1.5%. TSMC is reportedly sold out until 2028. And agentic AI is just getting started, along with using AI for real productivity (not just for getting directions to the mall).

Every dollar of buildout inflation moves from the companies that have to spend to the companies that get paid. Hyperscalers can’t slow down, because nobody wants to fall behind in this race. So the checks keep getting bigger, and more of them get paid with debt and new stock that dilutes shareholders. That’s where the money bottleneck shows up, so I’m watching credit spreads. The companies with pricing power and order books running past 2030 are the ones collecting.

Rising costs show you who’s getting paid in the AI trade, and that’s where I want to be.

Here for a good time… AND a long time,

Hans