The artificial intelligence revolution is officially upon us. If the abrupt improvement in your co-worker’s email grammar didn’t tip you off, the drastic increase in your power bill is a hard-to-miss clue. (And if your bill hasn’t changed much yet, consider yourself lucky.)

As the massive, power-hungry data centers that power AI’s expansion come online, they’re competing for power with everyone else. That’s driving up energy bills for industry and consumers alike, while testing the limits of US energy production capacity and stressing an aging power grid.

Silicon Valley has a solution: nuclear energy. Big Tech is investing heavily in the long-shunned (in the US, at least) energy source to power its AI moment, mostly in the form of so-called small nuclear reactors (SMRs), the next-gen version of nuclear tech that can (theoretically) be mass-produced and strategically deployed. (For the uninitiated, it might be helpful to think of SMRs as gas generators on radioactive steroids.) Now, a nuclear boom is directly downstream from the AI boom: According to a recent Bloomberg Intelligence report, soaring power demand from AI will spur $350 billion in nuclear spending in the US by 2050.

The US government, which views dominance in the AI sphere as crucial to continued economic and geopolitical dominance around the globe, is entirely on board. In a rare instance of bipartisan consensus, both the current and previous administrations have moved fast to cut red tape, overhaul oversight processes, and pour capital into the resurgent nuclear industry.

“We’re in a very serious bind. We’ve already tapped out traditional oil and gas technologies. There’s an eight-, nine-year queue for diesel generators, the most expensive form of energy, and now also gas turbines,” Kevin Kong, founder and CEO of AI-driven nuclear compliance platform Everstar, told The Daily Upside. “Renewables are not dense enough … Data centers run 24/7, and are extremely power-dense. And so the only technology that’s left that was overlooked and under-invested in is nuclear.”

In other words, if we’re going to have an AI revolution, we’ll need plenty of nukes. But will the industry insulate Americans from rising energy bills? Maybe.

The Troubling Data on Data Centers

For years, experts had estimated energy demand growth in the years and decades to come based on banal drivers such as population growth, economic expansion and development of emerging economies, as well as the electrification of everything, including major industries like manufacturing and transportation. It would be predictable and hence manageable, they believed.

Then came ChatGPT. Now? Most estimates predict that global energy demand will nearly double by 2050. A recent report from the International Energy Agency (IEA) found that over 50% of that growth will be driven by AI expansion. For example, a ChatGPT query requires roughly 10 times the energy, on average, needed for a Google search. To put it in even starker perspective, the IEA estimates that a typical AI-focused data center consumes as much electricity as 100,000 homes, while the largest such data centers consume 20 times that amount.

According to a recent Goldman Sachs report, data center power demand is expected to increase 160% by 2030 alone, and meeting 60% of that demand will require new energy generation capacity. Meanwhile, a report produced by Lawrence Berkeley National Laboratory and published by the Department of Energy estimates that data centers will consume more than 12% of total US electricity by 2028, up from 4.4% in 2023.

The triple-digit growth figures are already having a triple-digit impact on those suddenly, and sometimes unwillingly, competing with data centers for electricity. According to a recent Bloomberg analysis of energy data, monthly electricity costs in areas near data centers are now 267% higher than just five years ago, at the dawn of the AI age (that compares with a cumulative overall inflation rate of about 25%).

“Without mitigation, the data centers sucking up all the load is going to make things really expensive for the rest of Americans,” said David Crane, chief executive officer of Generate Capital.

Sweet Relief? The good news is that mitigation is coming. That $350 billion spending boom estimated by Bloomberg is also projected to increase the US’s nuclear energy output by 63%. And a rise in nuclear energy production has been proven to lower energy costs. Consider France, where a recent nuclear boom means the energy source now accounts for roughly 67% of total electricity supply, according to the 2025 World Nuclear Industry Status Report (that compares with just under 20% in the US). The chain reaction? Average day-ahead power prices fell to the lowest levels for the month of September in two decades in Epex Spot SE auctions. A similar story can be seen in South Korea, where nuclear energy now accounts for more than 30% of power.

“About 60% of the grid should be coming from nuclear; replace all natural gas and coal with nuclear. That’s the goal: Make the United States look like France,” Kong said.

Which brings us to the bad news. That 63% increase isn’t expected to be achieved until … 2050. In other words, mitigation is still years away. And the worse news: Relative to peers, the US is really, really bad at developing and deploying nuclear energy. When the Plant Vogtle nuclear power plant’s third reactor came online in 2023 in Burke County, Georgia, it marked the first from-scratch new nuclear reactor in the US in more than three decades. Infamously, the project came in seven years over schedule and some $17 billion over budget, or more than double initial budget projections, with the delays driving Westinghouse Electric into bankruptcy.

There are currently no nuclear reactors under construction in the US, per the Industry Status Report, putting it well behind peer nations. China, for instance, has only 59 reactors online (compared with 94 in the US), but has 33 more under construction. That’s roughly half of all nuclear reactors being built around the globe. By 2030, according to Goldman Sachs, the nation will top the US as the world’s leading nuclear energy producer.

In more grim news for Washington, the US has failed, since its Cold War-era nuclear heyday, to generate the next generation of nuclear experts. It produced just 929 graduates in nuclear engineering in 2022, down 20% from a peak in 2015. Per Goldman Sachs, the US will need to triple its US-based nuclear workforce to meet demand by 2050 (Kong says AI platforms like Everstar can help fill the knowledge gaps in the meantime).

Some more good news: SMRs, theoretically, will be much quicker to build and deploy than giant, Vogtle-style reactors.

“It takes 10 years, in the best case, to build that gigawatt-scale plant,” Matt Loszak, co-founder and CEO of SMR start-up Aalo Atomics, told The Daily Upside. “But if you mass manufacture [SMRs] in a factory, and you do more in parallel, you can actually deploy [SMRs] in a matter of a year or two instead of 10 years.”

Bipartisan consensus on cutting red tape for nuclear projects may speed up the process, too. Loszak says Aalo, which recently scored a $100 million Series B funding round and is one of 10 companies participating in the Department of Energy’s test reactor program, has benefited from regulatory frameworks streamlined by both the Trump and Biden administrations. That means actual deployment can come by the start of the next decade.

“The whole goal here is by 2030 to have a factory set up that can actually output at least a gigawatt per year of capacity,” Loszak says. (A single Aalo “Pod” SMR generates 50 megawatts of capacity; there are 1,000 megawatts in a gigawatt.)

One Bubble After Another: While SMRs are likely to deliver consistent energy supplies to massive data centers, a valuable proposition in its own right, not everyone is convinced it will be delivered cheaply.

According to data from Wood Mackenzie recently seen by the Financial Times, the “levelised cost of energy” for SMRs, or the cost for power that should be charged for the project to break even, will be around $182 per megawatt hour in 2030. That compares to $133 per megawatt-hour from traditional nuclear power plants, such as Vogtle, $126 for natural gas, and even less for wind and solar.

“SMRs are going to help with the hyperscalers,” who will be willing to pay a premium for 24/7 access to behind-the-meter energy, Kong said. But that might result in a bottleneck and misallocation of time, energy and resources toward SMRs rather than the type of large-scale, grid-changing power plants that have helped lower energy costs in countries around the world.

“There’s a time and place for SMRs,” Kong said. “But if we say we want a cleaner form of energy for hospitals, for you and me every day, for turning on the kettle and for electric vehicles that are becoming more and more prevalent, then it really has to come from the gigawatt-class reactors.”