Why AI’s future might depend on nuclear energy

Every year in late April, Earth Day makes me think about energy. This year, it also made me remember what Sam Altman recently said about artificial intelligence (AI) and energy—how we haven’t really grasped how much energy AI will need, and that we’ll need some kind of breakthrough in how we generate electricity to support the technology’s widespread use. He should know. OpenAI’s ChatGPT probably uses about 500,000 kilowatt-hours (kWh) of electricity PER DAY—as much as 17,241 average US households. Eventually, AI may consume as much energy as a country the size of the Netherlands. Rene Haas, CEO of chip-design company Arm, recently summed it up: AI’s energy needs are not sustainable given our current infrastructure.

Globally, utilities are already struggling to keep up as hotter temperatures increase the need for cooling, more electric vehicles increase the need for charging, and all of these awesome hyperscaler data centers increase the need for 24/7 power. Even with chipmakers continuing to deliver big efficiency gains and Saudi Arabia’s effort to build the world’s biggest solar power plant, adding AI’s huge energy appetite to the mix means we need to find new, reliable, and low-carbon sources of energy. And fast.

In the same speech I quoted above, Sam Altman said that he wished the world would embrace nuclear power. In fact, he’s putting his money where his mouth is, investing in a couple of different nuclear startups: one fission, and one fusion. Elon Musk has a similar view, saying that until we have more renewable power in place, we have to keep nuclear power. And both Microsoft and Amazon are also planning to use nuclear power for their data centers in the Eastern US, reflecting a growing corporate trust of nuclear energy.

There’s a lot to like about nuclear: reliability, scalability, and a low carbon footprint. But it also has plenty of challenges, including safety issues, the upfront cost, and what to do with the waste.

So, is nuclear power is the answer?

The benefits

A nuclear reactor can provide a massive amount of scalable, always-on power. Just ask France, which gets the majority of its electricity from its fleet of nuclear plants. Recently, nuclear power’s small carbon footprint compared to fossil fuels has brought the technology a new set of enthusiastic fans: environmentalists.

Reliable, 24/7 power supply

Nuclear power’s biggest advantage is its ability to provide the consistent, large-scale electricity needed to run energy-hungry AI applications. Unlike solar and wind, nuclear power is still going strong after the sun sets and the wind dies down. It has a predictable, stable output that can keep the digital engines of AI humming around the clock.

Low carbon footprint

As the world grapples with the existential threat of global warming, nuclear power’s relatively low emissions profile is another big plus. Powering AI with nuclear rather than fossil fuels can significantly reduce the environmental impact of this rapidly growing technology.

Scalability and flexibility

Because the fuel is compact and plant design is modular, nuclear reactors offer the flexibility to deploy power generation in increments. That makes it easier to scale up or down to meet changing demands, and allows nuclear energy to be integrated with other sources in hybrid systems to boost reliability and sustainability.

The challenges

Of course, nuclear power comes with some significant baggage. The specter of disasters at Three Mile Island, Chernobyl, and Fukushima Daiichi loom large in the public consciousness. Other drawbacks include the high upfront costs, slow pace of deployment, issue of waste storage, and shortage of trained professionals needed to operate the facilities.

Safety concerns

The headline, when we talk about the safety of nuclear energy, is about the potential for catastrophic accidents that would release unsafe levels of radiation into the environment. We can’t ignore vulnerability to malicious attacks due to terrorism and acts of war, although recent advances in nuclear technology have improved efficiency and safety.

Radioactive waste

Spent nuclear fuel must be safely contained for several thousand years. Even though modern technology produces less waste, any amount of it is expensive and difficult to transport and store, and nobody wants it in their backyard. There’s still not a good answer to this problem.

High upfront costs and slow deployment

Building new nuclear plants is incredibly expensive, with costs that can be three to four times higher than solar or wind per megawatt-hour. At the same time, the heavily regulated nature of the industry means it takes years to get a new plant up and running. 

The path forward

Interestingly, there may be a symbiosis in the nuclear-AI pairing. The advanced analytics and automation capabilities of AI could help drive innovations that address nuclear’s challenges, from enhancing safety to optimizing operations. So, the two technologies may end up in a mutually beneficial relationship, each helping the other overcome its shortcomings.

Yet ultimately, as AI transforms every industry, the need for reliable, sustainable power will only grow, and the need to contain carbon emissions will become more urgent. Here’s my take: until we reach a future where all our energy sources are clean, green, and sustainable, nuclear power is a critical part of the energy mix. It’s more expensive and dangerous than solar and wind, but it provides the 24/7 power generation that those renewables can’t, and is a ready source of low-carbon energy, which we desperately need right now. Where we already have plants, we should keep them going. The best option in the electricity game: all of the above, nuclear included.

Recent Posts

Get my FREE insider newsletter, delivered every two weeks, with curated content to help telco execs across the globe move to the public cloud.

"*" indicates required fields

Email*

Comments

This field is for validation purposes and should be left unchanged.