Nuclear's Unlikely Savior

Let me tell you about the most unexpected plot twist of 2024-2026.

The technology industry—the same industry that prides itself on being "green," that plasters solar panels on every roof, that publishes glowing ESG reports—just became the world's biggest advocate for nuclear power.

Not because they suddenly care about carbon emissions.

Because they need electricity. Massive amounts of it. 24/7, 365 days a year. And nothing else can deliver it.

Here are the facts:

Ten years ago, nuclear was dead. Fukushima had happened. Germany was shutting down reactors. New projects were billions over budget and years behind schedule. Public opinion was toxic.

Today? Nuclear is the hottest infrastructure bet in the world.

And it's not governments driving this renaissance. It's Big Tech.

Because when you need to power the AI revolution, and solar doesn't work at night, and wind doesn't blow on demand, and the grid can't handle your load—you go nuclear.

This is the story of how AI just saved nuclear power. And how nuclear might be the only thing that can save AI.

Let's start with why tech companies are doing this.

AI data centers have a problem that most people don't understand: they need constant, uninterrupted power.

Not "power most of the time." Not "power when the sun shines." Power every single second, at full capacity, forever.

Here's why:

This is called baseload power—electricity that's available 100% of the time, regardless of weather, time of day, or grid conditions.

And here's the problem: renewables can't provide it.

You can overbuild renewables and add massive battery storage. But the economics don't work at data center scale. A 300 MW data center running 24/7 would need:

And even then, you're vulnerable to multi-day weather events.

Nuclear, by contrast, has a capacity factor of 90-95%. It runs continuously for 18-24 months between refueling. It's carbon-free. And it can be built on-site, eliminating grid interconnection delays.

For AI data centers, nuclear isn't just an option. It's the only option that works at scale.

Let's talk about the deal that changed everything.

In September 2024, Microsoft signed a 20-year power purchase agreement with Constellation Energy to restart Three Mile Island Unit 1.

Yes, that Three Mile Island. Site of the 1979 partial meltdown that killed the U.S. nuclear industry for a generation.

The reactor will be renamed Crane Clean Energy Center (because branding matters). It will generate 835 MW of carbon-free electricity—enough to power a large data center campus. Microsoft gets guaranteed power. Constellation gets a guaranteed customer.

The deal is worth an estimated $16 billion over 20 years.

And here's the kicker: Unit 1 was never involved in the accident. It operated safely for decades and was only shut down in 2019 because it couldn't compete economically with cheap natural gas.

Now, with AI demand and Microsoft's money, it's coming back online.

The symbolism is hard to overstate. The reactor that symbolized nuclear's death is being resurrected to power the AI revolution.

But restarting old reactors is just the beginning. The real story is Small Modular Reactors (SMRs)—a new generation of nuclear technology designed specifically for the AI era.

Here's what makes SMRs different:

1. They're small.
Traditional nuclear plants generate 1,000+ MW. SMRs generate 50-300 MW—perfect for a single data center campus.

2. They're modular.
SMRs are built in factories and shipped to site, rather than custom-built on-site. This reduces construction time from 10-15 years to 3-5 years.

3. They're safer.
SMRs use passive safety systems that don't require electricity or human intervention. If something goes wrong, physics shuts the reactor down automatically.

4. They're flexible.
You can deploy one SMR or ten, depending on your power needs. They can be sited almost anywhere—including directly at data center campuses.

And tech companies are going all-in:

Amazon announced three SMR projects in October 2024:

Google signed a deal with Kairos Power to purchase electricity from multiple SMRs totaling 500 MW, with the first reactor online by 2030.

Oracle is designing a 1 GW data center powered by three SMRs—the largest SMR-powered data center ever proposed.

Total announced SMR capacity for data centers? Over 2 GW and growing.

You might be wondering: if SMRs are so great, why didn't we build them a decade ago?

Three reasons:

1. Economics didn't work.
Without a guaranteed customer willing to pay premium prices for 24/7 carbon-free power, SMRs couldn't compete with cheap natural gas or subsidized renewables.

2. Regulatory uncertainty.
The U.S. Nuclear Regulatory Commission (NRC) had never licensed an SMR. Developers faced years of uncertainty and potential delays.

3. Public opposition.
Post-Fukushima, nuclear was politically toxic. No utility wanted to be the first to try.

What changed?

1. AI created desperate demand for baseload power.
Tech companies are willing to pay $100-150/MWh for guaranteed 24/7 electricity—far above market rates. That makes SMR economics work.

2. The NRC approved the first SMR design.
In January 2023, the NRC issued the first-ever design certification for an SMR (NuScale's design). This created a regulatory pathway for other designs.

3. Climate commitments made nuclear politically acceptable.
Tech companies have net-zero carbon pledges. Nuclear is the only carbon-free baseload option. Suddenly, the politics shifted.

4. Grid interconnection queues became untenable.
When connecting to the grid takes 5-7 years, building your own power plant starts to make sense—even if it's nuclear.

The confluence of these factors created a once-in-a-generation opportunity for nuclear.

And tech companies—facing an existential need for power—seized it.

Here's where my framework gets interesting.

For 50 years, nuclear was scarce:

Meanwhile, natural gas and renewables were abundant:

AI just flipped the script.

Now:

And suddenly:

The scarcity shifted. And capital follows scarcity.

This is why nuclear stocks have surged:

Investors see the shift. Nuclear went from scarce demand to scarce supply overnight.

Let me be clear: this nuclear renaissance is not a sure thing.

Here are the risks:

1. SMRs have never been built at scale.
Every SMR project is essentially a prototype. We don't know if they'll actually cost what developers claim, or if they'll be delivered on schedule. History suggests: probably not.

2. Regulatory delays.
Even with NRC design certification, each individual project needs site-specific licensing. That process can take 3-5 years—and that's if everything goes smoothly.

3. Public opposition.
Not everyone is thrilled about nuclear reactors next to data centers. Expect lawsuits, permitting battles, and NIMBY opposition—especially in blue states.

4. Uranium supply constraints.
If hundreds of new SMRs come online, uranium demand will surge. Russia controls ~35% of global uranium enrichment capacity. Geopolitical disruptions could create supply shocks.

5. Cost overruns.
Nuclear projects have a notorious history of going over budget. The Vogtle nuclear plant in Georgia was $17 billion over budget and 7 years late. SMRs are supposed to avoid this with factory construction—but we won't know until the first few are built.

6. Technological competition.
What if fusion power becomes commercially viable in 10 years? What if next-gen geothermal scales faster than expected? SMRs could become obsolete before they're even deployed.

These are real risks. And anyone investing in nuclear needs to understand them.

Winners:

1. Existing nuclear operators (Constellation, Exelon, Duke Energy)
Their existing reactors just became strategic assets. Expect life extensions, uprates, and premium power contracts.

2. SMR developers (NuScale, Kairos Power, X-energy, TerraPower)
If even 10% of announced projects get built, these companies will be worth tens of billions.

3. Uranium miners (Cameco, Kazatomprom, Energy Fuels)
Uranium demand is set to surge. Prices have already doubled from 2020 lows.

4. Nuclear fuel cycle companies (Centrus Energy, Urenco)
Enrichment, conversion, and fuel fabrication are all bottlenecks. Capacity is scarce.

5. Tech companies with early nuclear deals
Microsoft, Amazon, Google, Oracle—they locked in power supply before prices spike. Competitive moat.

Losers:

1. Renewable developers
Tech companies were supposed to be their biggest customers. Now that capital is flowing to nuclear instead.

2. Natural gas generators
Nuclear takes market share from gas peaker plants and combined-cycle generators.

3. Grid operators in anti-nuclear states
If data centers go off-grid with SMRs, states like California and New York lose economic development.

4. Late-movers in AI infrastructure
If you don't have power secured, you can't build data centers. Nuclear deals give first-movers a 5-10 year advantage.

Nuclear's comeback will reshape more than just the power sector.

1. Geopolitical shifts in uranium supply
Expect U.S. investment in domestic uranium mining and enrichment to reduce dependence on Russia and Kazakhstan.

2. Regulatory reform
The NRC will face pressure to streamline licensing for SMRs. Expect faster approvals and standardized processes.

3. Nuclear talent shortage
The U.S. hasn't built new reactors at scale in 30 years. There's a critical shortage of nuclear engineers, operators, and construction workers. Universities are scrambling to rebuild nuclear programs.

4. Climate policy realignment
Environmental groups are split. Some embrace nuclear as essential for decarbonization. Others remain opposed. Expect fractures in the climate coalition.

5. State-level competition
Pro-nuclear states (Georgia, Tennessee, Texas) will attract data center investment. Anti-nuclear states (California, New York) will lose out—unless they change course.

If you're an investor:

If you're building AI infrastructure:

If you're in policy:

If you're an environmentalist:

If you're a citizen:

Let me end with a story that crystallizes everything.

In March 2026, Italy—a country that voted to ban nuclear power in a 1987 referendum after Chernobyl—announced it's exploring a return to nuclear energy.

Why? Energy costs and AI demand.

Italy has some of the highest electricity prices in Europe (€0.30-0.40/kWh vs. €0.15-0.20 in France). Italian data centers can't compete. AI companies are building in France instead—which gets 70% of its electricity from nuclear.

Italy's choice: watch the AI economy happen elsewhere, or reconsider nuclear.

They're reconsidering.

This is the pattern we'll see globally:

Countries with cheap, reliable electricity attract AI investment.
Countries without it lose out.

And the only proven technology that delivers cheap, reliable, carbon-free electricity at scale is nuclear.

So countries will face a choice:

Germany is already having this debate. Japan is restarting reactors. South Korea is reversing its nuclear phase-out.

AI is forcing a global nuclear renaissance.

Not because of climate commitments or energy security (though those help).

Because of economics.

The countries and companies that secure nuclear power will dominate the AI era.

The ones that don't will watch from the sidelines.

Here's what happened:

For 50 years, nuclear was dying. Too expensive. Too slow. Too politically toxic.

Then AI came along and said: "We need 24/7 carbon-free power at gigawatt scale, and we need it in 5 years, and we'll pay whatever it costs."

And suddenly, nuclear was back.

Not because of government policy. Not because of climate activists. Because of Microsoft, Amazon, and Google writing billion-dollar checks.

The tech industry—accidentally, ironically—just became nuclear power's savior.

And nuclear—improbably, unexpectedly—might be the only thing that can power the AI revolution at scale.

The math is simple:

So we're going to see:

This is a turning point.

Not just for energy. For geopolitics, climate policy, industrial strategy, and the future of AI itself.

Because here's the deeper truth:

AI is not just a software revolution. It's an energy revolution.

And energy revolutions don't happen in the cloud. They happen in the physical world—with atoms, not bits.

We're building:

That's $5-7 trillion in physical infrastructure to support a technology that most people think of as purely digital.

The atoms matter.

And the companies, countries, and investors who understand this—who see that AI is fundamentally a materials and energy story, not just a software story—will be the ones who win the next decade.

Nuclear was dead.

AI brought it back to life.

And now, nuclear might be the only thing that can keep AI alive.

The irony is almost too perfect.

But that's how abundance and scarcity work. The bottleneck shifts. The scarce becomes abundant. The abundant becomes scarce.

And the world rearranges itself around the new reality.

Welcome to the nuclear age. Again.

Primary Sources (Accessed March 2026):

Data Sources:

Additional Reading:

Next in the series: Part 4 - "The Inference Economy: Why Running AI Models Will Cost More Than Training Them"

Follow @SloneSterling on X.com for daily research on AI, energy, commodities, and the collision of abundance and scarcity.

Precision in a world of noise.

Analysis by Slone Sterling