CEG might be the cleanest AI nuclear stock. The valuation is the hard part.

I’ve been looking at the nuclear names because the AI trade seems to be spreading from GPUs to memory, optics, and now power.

CEG is probably the cleanest single-stock version of that nuclear power thesis.

Not the most exciting one. Not the most memeable one. But maybe the cleanest.

The reason is pretty simple: Constellation already owns real nuclear assets and already generates real cash flow. That matters. A lot of the SMR names are still mostly a bet on future projects, future approvals, future financing, and future commercial operation. CEG is not that.

That does not automatically make it cheap though.

Based on quote data I pulled on May 20, CEG was around $260.67, with a market cap around $94B. The stock is down roughly 35% from its 52-week high, so it no longer looks like the straight-line AI power winner it looked like a few months ago.

But it still trades around 19x forward earnings and about 14.6x EV/EBITDA. That is not exactly sleepy utility pricing.

That is the whole tension here.

The bull case is easy to understand. Data centers need power that is reliable, large-scale, and increasingly carbon-free. Nuclear fits that better than most sources. The EIA made the point pretty plainly back in 2024: data centers do not consume electricity like normal households. They need steady power around the clock, and nuclear plants are built for that kind of constant output.

CEG also has actual hyperscaler validation. Constellation signed a 20-year power purchase agreement with Microsoft to restart Three Mile Island Unit 1 as the Crane Clean Energy Center. The company said that deal would add around 835 MW of carbon-free power to the grid and support Microsoft’s data center power matching goals.

Meta is part of the same broader story. In January, Meta said its nuclear agreements with Vistra, TerraPower, Oklo, and the prior Constellation deal made it one of the biggest corporate nuclear power buyers in U.S. history. That is a pretty big signal that this is not just a retail stock-market theme.

So yes, the thesis is real.

The problem is that real thesis and good entry are not the same thing. I keep coming back to that with CEG.

If you compare it with the other AI power names, CEG sits in a weird middle. It is much more underwritable than OKLO or SMR because it has operating assets and earnings. But it is also not obviously cheap versus VST or TLN, both of which screen cheaper on forward earnings in the Yahoo data I pulled.

CEG looks like the quality name. The market knows that.

The question is whether investors are paying for nuclear assets, or paying for a nuclear scarcity story.

Those are different trades.

If you are paying for the existing fleet, then valuation discipline matters. You want earnings growth, contract visibility, buybacks, and operating execution. Boring stuff. Useful stuff.

If you are paying for scarcity, then the stock needs more than normal utility results. It needs more hyperscaler contracts, more evidence that AI load growth is tightening power markets, and a clean path for projects like Crane to come online without the usual regulatory/grid delays.

That is where I get a little cautious.

The Microsoft deal is great, but the restart still depends on regulatory approvals and project execution. EIA also pointed out that data center demand has uncertainty around how much capacity actually gets built, how quickly those sites reach peak power use, and how efficiency changes over time.

In other words, “AI needs power” is true. But it is not specific enough by itself to value CEG.

The setup I see is more like this:

CEG is probably one of the best ways to express the AI nuclear theme if you want actual assets and not just a science project. But after the rerating, the stock needs the AI power story to keep turning into signed contracts and earnings, not just vibes.

Strong company. Real theme. Not an obvious bargain.

If the stock were trading like a normal utility, I think the argument would be easier. At roughly 19x forward earnings, I would put it more in watchlist territory unless someone believes the market is still underpricing the value of contracted nuclear power for data centers.

The thing I would watch is pretty specific: do the next few catalysts look like real contracted cash flow, or just more headlines about AI electricity demand?

That is probably the line for me.

Curious how others are thinking about CEG versus the higher-beta nuclear names. Would you rather own the existing nuclear fleet with a premium multiple, or take the messier upside in OKLO / SMR-type names?