Hey Income Traders,
There’s a simple way to think about this whole AI buildout.
Forget models for a second. Forget chatbots. Forget the apps.
Start with one question: How do you move all that power?
Because AI doesn’t float in the cloud. It sits in data centers. And data centers don’t run on ideas. They run on electricity.
How much electricity? U.S. data center power demand hit 62 gigawatts in 2025. By 2030, it triples to 134 gigawatts. For context, that’s like adding ten New York Cities to the grid, just for servers.
Which means someone has to build the lines, upgrade the grid, install transformers, rewire old systems, and move massive amounts of current from point A to point B.
And that means copper. Lots of it.
Microsoft’s Chicago data center used 2,177 tonnes of copper. That’s 27 tonnes per megawatt. A single AI-ready site running at 100 megawatts locks in 3,000 tonnes before you even count the grid upgrades feeding it.
Every new data center, every new substation, every new transmission upgrade is basically a copper delivery system with a roof on it.
This is the part of the story most people skip because it isn’t flashy.
No avatars. No agents. No demos.
Just trenches, pylons, and wire.
But that’s where the constraint lives.
You can announce a thousand new AI projects. But if the grid can’t carry the load, none of them matter. Two data centers in Silicon Valley sit finished right now, waiting to turn on. They can’t. The transformers aren’t available. Lead times for critical power equipment hit 144 weeks. Wood Mackenzie says we’re 30 percent short of what we need in 2025 alone.
That’s why the companies that quietly design, build, and maintain the power backbone keep showing up in my work. They’re not predicting the future. They’re being forced into it.
Follow the Checks
Utilities have to upgrade. Data centers have to connect. Governments have to harden infrastructure. And none of that happens without specialists who already know how to do it.
The checks are already clearing. U.S. utilities will spend $1.4 trillion on grid infrastructure from 2025 to 2030. That’s double the prior decade. MISO alone just greenlit $21.8 billion in new high-voltage lines. Texas approved $13 billion more.
This isn’t cyclical demand. It’s structural.
S&P Global just called the copper supply gap “systemic risk.” Copper supply peaks in 2030. Demand rises 50 percent by 2040. The shortfall: 10 million metric tons. That’s not a dip. That’s a decade-long hole.
You don’t pause grid upgrades once workloads arrive. You don’t roll back copper once lines are laid. You don’t mothball substations because inflation ticked down.
Once the build starts, it tends to keep going.
That’s what makes this theme so different from software stories. Software competes. Infrastructure compounds.
What Comes Next
I’ll send details soon, but here’s the punchline:
If AI is the demand shock, the grid is the battlefield.
And copper is the ammo.
May the income be with you,
Hans