AI Money Is Moving Beyond Chips — Why SK Is Looking at US Nuclear

SK Group is reviewing an investment in a US nuclear power company. Reports suggest the scale could run into trillions of won. Why is a chip conglomerate looking at power plants?

Electricity. AI consumes an enormous amount of it.

How much power does AI actually need?

Training and running AI models happens entirely inside data centers. Tens of thousands of accelerator cards run around the clock, and the cooling equipment that keeps them from overheating draws power of its own. A single large data center can consume as much electricity as a small city.

That's why big tech companies started signing deals directly with power producers a few years ago. They aren't just buying electricity — they're securing the generating capacity itself. Solar and wind only deliver when the weather cooperates, which is how nuclear ended up on the shortlist.

Follow where the money goes

It went to AI chipmakers like Nvidia first. Memory came next — the high-bandwidth memory stacked beside those chips ran short, and SK Hynix and Samsung benefited. But you can manufacture all the chips you like; without data centers to install them in, they do nothing.

So data centers are the next square: buildings, cooling systems, and power. Once you reach power, the chain keeps going — generators, transmission lines, transformers, cable makers. SK looking at a US nuclear company is an attempt to stand at the far end of that chain, where the electricity is actually made.

What's striking is how long this could run. Power plants take years to build. Chip output can be scaled by running fabs harder; electricity cannot be scaled that quickly. In any market, the side that is slow to add supply tends to stay expensive the longest.

Worth staying careful about

Plenty of people think AI demand is overstated. Nobody can promise that the data centers under construction today will be full in a few years. Power investments have long payback periods, so a wrong demand forecast leaves losses that linger. And SK's investment is still only under review.

A semiconductor group studying power plants would have sounded odd a few years ago. Today it reads like the obvious next step.