Every AI Data Center Needs This Company First. The Market Hasn't Done the Math.

Two years ago this was a sleepy utility contractor. Now it's the chokepoint of the AI economy.

Here’s a number that should stop you mid-scroll: 160 weeks.

That’s how long you wait — today — to get a substation transformer in America. Three years. For a piece of equipment that every single AI data center needs before a single GPU powers on.

Now hold that thought next to this one: hyperscalers — Microsoft, Amazon, Google, Meta and friends — confirmed roughly $725 billion in capital spending for 2026, with about $545 billion of it aimed squarely at AI infrastructure. Data center construction starts hit $49.5 billion through April — versus $13.6 billion in the same stretch last year. That’s not growth. That’s a 3.6x land rush.

So we have the largest infrastructure buildout since the interstate highway system... slamming into a grid that physically cannot keep up.

When unstoppable demand meets an immovable bottleneck, somebody gets pricing power. And it’s usually not who you’d guess.

It’s not the chipmakers — everyone owns those. It’s not the hyperscalers — they’re the ones writing the checks. It’s the companies that control the scarce thing: the crews, the cranes, and the utility relationships that actually connect data centers to power.

There’s a century-old company sitting directly on this chokepoint. It just reported record revenue, record profit, and a record backlog — and it’s quietly consolidating a few percent below its all-time high while Wall Street obsesses over the Mag-7.

Most investors have never heard of it. The ones who have can’t buy enough of it.

But here’s where the story gets interesting... because the market may be completely mispricing what happens next.