AI’s next bottleneck is the power bill

Not every AI winner wears a hoodie

Nvidia turned the AI trade into a blockbuster, but this piece argues the next leg of the story may be less about chips and more about the plumbing. Bank of America’s big number is the kind of thing that makes your jaw drop: a potential $1.7 trillion AI data-center systems market by 2030, with capacity doubling and trillions in capex floating around like confetti.

The real constraint? Electricity, not enthusiasm

Here’s the awkward part for all that AI optimism: you can’t run giant server farms on vibes alone. The article points to power, labor, and supply chains as the actual bottlenecks, and even says U.S. data-center demand could leave utilities staring at a more than 100-GW generation shortfall through 2030. Translation: the AI boom may be bottlenecked by transformers, switchgear, cooling, and enough electrons to keep the whole circus humming.

ETFs are taking the scenic route

Instead of betting on a handful of mega-cap tech names, the article pitches infrastructure ETFs as a way to play the second-order winners. GRID gets the spotlight, with holdings like Quanta Services, Eaton, Johnson Controls, ABB, and Schneider Electric. PAVE and IFRA show up too, giving investors broader exposure to the construction, electrification, and industrial side of the AI buildout.

Big picture

Caterpillar also gets a nod because its power-generation equipment is showing up in data-center applications. The message is pretty simple: if AI is the hype engine, power infrastructure may be the gas station. And in a market obsessed with the next shiny model, the boring stuff could still be where the money quietly stacks up.