Going nuclear will be the only way to keep the lights on as AI guzzles ever more electricity

What struck me wasn’t the intelligence of this machine. It was the sheer energy it was devouring. The engineer said, “This one building consumes as much power as a small town.” That’s when the magnitude of the challenge hit me: If AI is our future, how on earth will we power it?

Also Read: AI as infrastructure: India must develop the right tech

All that intelligence takes energy. A lot of it. More than most people realize. And as someone who’s spent years studying the physics of energy systems, I believe we are about to hit a hard wall. To be blunt: AI is growing faster than our ability to power it. And unless we confront this, the very tools meant to build our future could destabilize our energy systems—or drag us backward on climate. 

One solution has been pinpointed by the AI industry: nuclear energy. Most people don’t associate AI with power plants. But every chatbot and image generator is backed by vast data centres full of servers, fans and GPUs running day and night. These machines don’t sip power. They guzzle it. In 2022, data centres worldwide consumed around 460 terawatt-hours. 

But that’s just the baseline. Goldman Sachs projects that by 2030, AI data centres will use 165% more electricity than they did in 2023. And it’s not just about scale. It’s about reliability. AI workloads can’t wait for the sun to shine or wind to blow. They need round-the-clock electricity, without fluctuations or outages. That rules out intermittent renewables for a large share of the load—at least for now. 

Also Read: Rely on modern geothermal energy to power our AI ambitions

Can power grids handle it?: The short answer: not without big changes. In the US, energy planners are already bracing for strain. States like Virginia and Georgia are seeing huge surges in electricity demand from tech campuses. One recent report estimated that by 2028, America will need 56 gigawatts of new power generation capacity just for data centres. That’s equivalent to building 40 new large power plants in less than four years. 

The irony? AI is often promoted as a solution to climate change. But without clean and scalable energy, its growth could have the opposite effect. For example, Google’s carbon emissions rose 51% from 2019 to 2024 by its own assessment, largely on account of AI’s appetite for power. This is an infrastructure emergency. 

Enter nuclear energy—long seen as a relic of the Cold War or a post-Chernobyl nightmare. But in a world hungry for carbon-free baseload power, nuclear power is making a quiet comeback. 

Let’s be clear: nuclear energy is the only scalable source of clean electricity in existence that runs 24/7. A single large reactor can power multiple data centres without emitting carbon or depending on weather conditions. 

Also Read: India should keep all its nuclear power options in play

Tech companies are already acting: Microsoft signed a deal to reopen part of the Three Mile Island nuclear plant to power its AI operations. Google is investing in small modular reactors (SMRs). These are compact next-generation nuclear units that are designed to be safer, faster to build and considered ideal for campuses. They’re early signs of a strategic shift: AI companies are realizing that if they want to build the future, they’ll have to power it themselves.

As a physicist, I’ve always been fascinated by nuclear energy’s elegance. A single uranium pellet—smaller than a fingertip—holds the same energy as a tonne of coal. The energy density is unmatched. But it’s not just about big reactors anymore. 

The excitement stems from advanced reactors. SMRs can be built in factories, shipped by truck and installed near tech campuses or even remote towns. Molten salt reactors and micro-reactors promise even greater safety and efficiency, with lower waste. New materials and AI-assisted monitoring make this technology far safer than past generations. For the first time in decades, nuclear power is both viable and vital. 

Also Read: Sumant Sinha: Nuclear energy is a crucial piece in the puzzle of climate action

But let’s talk about the risks: I’m not naïve. Nuclear still carries a stigma—and poses real challenges. Take cost and time; building or reviving reactors takes years and billions of dollars. Even Microsoft’s project will face regulatory hurdles. 

Or waste; we still need better systems for storing radioactive materials over the long-term. Or consider control; if tech giants start building private nuclear plants, will public utilities fall behind? Who gets priority during shortages? And of course, we must be vigilant about safety and non-proliferation. The last thing we want is a tech-driven nuclear revival that ignores the hard lessons of history. 

But here’s the bigger risk: doing nothing. Letting power demand explode while we rely on fossil fuels to catch up would be a disaster.

We live in strange times. Our brightest engineers are teaching machines to think. But they still haven’t solved how to power those machines sustainably. As a physicist, I believe we must act quickly—not just to make AI smarter, but to make its foundation stronger. Nuclear energy may not be perfect. But in the race to power our most powerful technology yet, it may be the smartest bet we’ve got. The AI revolution can’t run on good intentions. It will be run on electricity. But where will it come from?

The author is a theoretical physicist at the University of North Carolina at Chapel Hill, United States. He posts on X @NishantSahdev