Why use nuclear instead of solar or wind for AI?

AI clusters run continuously, so they need always-on power. Solar and wind are cheap but intermittent. Nuclear is the one proven source that is both clean and runs 24/7, which is why buyers who ignored it a decade ago now compete for every megawatt.

How do you invest in the nuclear and AI trend?

It is mainly a fuel-and-enabler story: uranium and the fuel cycle such as Cameco, enrichment and HALEU such as Centrus, the utilities that own the plants signing AI power deals, and next-generation or small modular reactor developers which carry the highest risk. Timelines are multi-year, so it is a structural bet rather than a quick trade.

The Lens · The Restart Wave

AI Is Reopening Dead Nuclear Plants

Q: Is AI really reopening nuclear plants?

Yes. The Crane Clean Energy Center, formerly Three Mile Island Unit 1 which shut in 2019, is restarting 835 MW targeted for 2027 under a long-term deal to power Microsoft's AI data centers. Hyperscalers are signing nuclear deals because AI needs clean, around-the-clock baseload power that solar and wind cannot guarantee on their own.

A nuclear reactor that shut down in 2019 is being brought back to life — and the customer it's restarting for is an AI data center. Meanwhile Canada just finished a massive reactor overhaul seven months early and under budget. The decade-long “nuclear is dead” consensus is reversing in real time, and the reason is simple: AI needs more clean, always-on power than the grid can build fast enough. Hover or tap any underlined term.

Dragonfly Lens · June 19, 2026 · Why AI's power hunger is reviving nuclear — and who profits.

The short version

What's real: the Crane Clean Energy Center (the old Three Mile Island Unit 1, shut in 2019) is restarting — 835 MW, targeted for 2027 — under a long-term deal to power Microsoft's AI buildout.
And in Canada: Bruce Power returned its Unit 3 reactor to service after a major refurbishment — ~7 months early and ~CAD$150M under budget, extending its life 30–35 years.
Why now: AI data centers need baseload clean power, and nuclear is the only proven source that delivers it around the clock.
The opportunity: it's a fuel and enabler story first — uranium, enrichment, the utilities that run the plants, and the next-gen reactor builders.
The honest line: these are multi-year timelines (2027+), restarts are rare (only a handful of candidates), and next-gen reactors are still unproven. Real trend, slow clock.

What's actually happening

For 20 years, nuclear plants closed and almost none opened. That just flipped, and AI is the catalyst:

A dead plant is being resurrected for AI. The Crane Clean Energy Center — formerly Three Mile Island Unit 1, shut down in 2019 as uneconomic — is being restarted to add 835 MW to the grid, targeted for 2027, under a 20-year deal tied to powering Microsoft's data centers. A plant that was uneconomic in 2019 is now worth reopening because AI changed the math.

Existing reactors are being life-extended, well. Canada's Bruce Power returned Unit 3 to service after a ~CAD$1.9bn refurbishment — seven months ahead of schedule and ~CAD$150M under budget — adding 30–35 years of life. (Note for the “nuclear always runs over budget” crowd: not this time.)

Hyperscalers are signing nuclear deals directly. Microsoft, Amazon, Google and Meta have all moved to lock up nuclear power — restarts, long-term contracts, and next-gen reactor bets — because it's the cleanest way to guarantee 24/7 power for AI.

Next-gen reactors are advancing. Small modular reactors (SMRs) and advanced designs — the OKLO-type bets — are pushing through licensing and fuel deals, aimed at the same AI-power demand a few years out.

Why now — the AI power math

The whole thing comes back to the power bottleneck. An AI data center is a power plant with chips attached — it draws enormous electricity, continuously. Solar and wind are cheap but intermittent; an AI cluster can't pause when a cloud passes. Nuclear is the one proven source that's both clean and always-on — which is exactly why the buyers who were indifferent to it a decade ago are now competing for every megawatt of it.

The honest catch: nuclear is the right answer on paper, but it's slow. Restarts take years; new builds take longer; SMRs are largely still pre-commercial. So nuclear doesn't solve AI's near-term power crunch — gas and the grid fill that gap. Nuclear is the multi-year clean-baseload bet underneath it, not the overnight fix.

Who profits

Layer	Who / how
Uranium & fuel	More running reactors = more fuel demand. The miners and fuel-cycle names (Cameco and peers) are the most direct exposure — the “sell shovels” layer of nuclear.
Enrichment & HALEU	Next-gen reactors need HALEU fuel, and domestic enrichment is a real bottleneck — a strategic chokepoint (Centrus and the enrichment chain).
The utilities that own the plants	Whoever owns the reactor signing the AI power deal (e.g. Constellation, the restart operator) captures the contracted revenue.
Next-gen / SMR developers	The advanced-reactor builders (the OKLO-type names) — highest upside, highest risk, furthest out. Speculative, label it so.
At risk: the “nuclear can't compete” thesis	The assumption that reactors stay uneconomic is the thing AI just broke.

The Lens angle

Don't trade the headline — trade the bottleneck. “Nuclear is back” is a vibe; the durable read is that AI's appetite for clean, always-on power is so large it's making uneconomic reactors economic again, on multi-year timelines. So the bet isn't the splashy SMR startup — it's the fuel and the enablers that get paid no matter which reactor wins: uranium, enrichment, the utilities holding the contracts. Same picks-and-shovels lesson as the rest of the AI buildout, and the same energy-is-the-ceiling truth underneath it.

The viral take and the true take are rarely the same trade

A dead reactor restarting for AI isn't a vibe. It's a power bottleneck with a uranium bill.

Dragonfly Lens maps where AI's power crunch actually flows — and who gets paid — while keeping the timelines honest. Plain English, every claim sourced.

Join the Lens →

More: The power bottleneck · The picks behind the picks · All explainers

Quick answers

Is AI really reopening nuclear plants? Yes. The Crane Clean Energy Center (the former Three Mile Island Unit 1, shut in 2019) is restarting 835 MW, targeted for 2027, under a long-term deal to power Microsoft's AI data centers. Hyperscalers are signing nuclear deals because AI needs clean, 24/7 baseload power that solar and wind can't guarantee alone.

Why nuclear instead of solar or wind for AI? AI clusters run continuously, so they need always-on power. Solar and wind are cheap but intermittent. Nuclear is the one proven source that's both clean and runs 24/7 — which is why buyers who ignored it a decade ago now compete for it.

How do you invest in the nuclear-AI trend? It's mainly a fuel-and-enabler story: uranium and the fuel cycle (Cameco and peers), enrichment/HALEU (Centrus and the enrichment chain), the utilities that own the plants signing AI power deals, and next-gen/SMR developers (highest risk). Timelines are multi-year, so it's a structural bet, not a quick trade.

Sources: Bruce-3 returned to service ~7 months early and ~CAD$150M under budget (~CAD$1.9bn refurbishment, 480 fuel channels / 960 end fittings / 8 steam generators replaced, +30–35yr life) — NucNet, World Nuclear News; Crane Clean Energy Center (ex-Three Mile Island Unit 1) 835 MW restart targeted 2027, Microsoft power deal — Interesting Engineering, Constellation disclosures.

Educational research, not personalized investment advice. Dragonfly Lens is not a registered investment advisor. Figures are as reported by the sources above. Company names illustrate the supply chain, not buy recommendations; nuclear timelines are long and projects carry regulatory and execution risk. Verify against primary sources before acting. Past performance does not guarantee future results.