Here’s the thing about the AI boom that nobody talks about at dinner parties: it’s not really one gold rush. It’s a sequence of gold rushes, each one solving a different problem before the next bottleneck shows up.
First, everyone needed GPUs. Nvidia printed money. Then hyperscalers realized they needed servers to hold those GPUs. Super Micro and Dell had their moment. Then the data centers started melting—literally—so cooling companies like Vertiv became the new hotness. Then the power grid started sweating, and suddenly nuclear energy was cool again (pun intended).
Each wave followed the same playbook: identify the constraint, throw billions at whoever solves it, watch their stock rocket.
We’re about to see it happen again. And this time, the bottleneck is hiding inside the data center itself.
The Problem Nobody’s Talking About
When you’re running hundreds of thousands of GPUs in a single cluster, the chips are only as fast as the cables connecting them. Seriously. If data can’t move between processors quickly enough, you’ve got expensive silicon sitting idle. That’s like owning a Ferrari but only driving it in a school zone.
This is where interconnects come in—the cables, switches, and chips that move data around inside the data center. It sounds boring. It’s not. It’s about to be huge.
Copper vs. Optics: The Technology Showdown
Here’s where it gets interesting. There’s a debate raging in the industry: should we use copper cables or fiber optics?
Copper is cheap, fast, and power-efficient. The problem? It only works for short distances. At today’s speeds, you’re looking at maybe three meters before the signal falls apart. It’s like trying to shout across a football field—works fine for 20 yards, but eventually people can’t hear you.
Optical fiber doesn’t have that problem. Light travels farther. But it requires active electronics, lasers, and photodetectors—all of which consume power and add cost.
The smart money says: copper wins now, optics wins later. There’s even a middle ground called active electrical cables that extends copper’s range while using less power than optics. It’s copper’s last stand before physics says nope.
The Stocks to Watch
If you want exposure to the copper phase (the next 2-3 years): Credo Technology (CRDO), Marvell Technology (MRVL), Broadcom (AVGO), and Amphenol (APH).
For the optical future (2027 and beyond): Lumentum (LITE), Coherent (COHR), and Fabrinet (FN).
The Bottom Line
The AI infrastructure boom isn’t ending. It’s just moving to the next problem. The hyperscalers have identified the bottleneck, and they’re about to deploy enormous capital to solve it.
The supply chain is constrained. The winners are already visible. And if history is any guide, the biggest gains go to investors who position themselves before the market figures it out.
That’s the real lesson of the AI boom: get there early, or get there late. There’s no in-between.