Why Big Tech is turning to nuclear to power its energy-intensive AI ambitions

Technology giants are turning to nuclear energy to power the energy-intensive data centers needed to train and run the massive artificial intelligence models behind today’s generative AI applications.

Microsoft and Google are among the firms agreeing deals to purchase nuclear power from certain suppliers in the U.S. to bring additional energy capacity online for its data centers.

This week, Google said it would purchase power from Kairos Power, a developer of small modular reactors, to help “deliver on the progress of AI.”

“The grid needs these kinds of clean, reliable sources of energy that can support the build out of these technologies,” Michael Terrell, senior director for energy and climate at Google, said on a call with reporters Monday.

“We feel like nuclear can play an important role in helping to meet our demand, and helping meet our demand cleanly, in a way that’s more around the clock.”

Google said its first nuclear reactor from Kairos Power would be online by 2030, with more reactors going live through 2035.

The tech giant isn’t the only firm looking to nuclear power to realize its AI ambitions. Last month, Microsoft signed a deal with U.S. energy firm Constellation to resurrect a defunct reactor at the Three Mile Island nuclear power plant in Pennsylvania, whose reactor has been dormant for five years.

The Three Mile Island plant was the location of the most serious nuclear meltdown and radiation leak in U.S. history in March 1979, when the loss of water coolant through a faulty valve caused a reactor to overheat.

Tech companies are under pressure to find energy sources to power data centers — a key piece of infrastructure behind modern-day cloud computing and AI applications.

Many developers rent out servers equipped with GPUs (graphics processing units), which would typically be too expensive to own outright, from so-called cloud “hyperscalers” — such as Amazon, Microsoft and Google.

These tech giants have benefited from a surge of interest in generative AI applications such as OpenAI’s ChatGPT. But that increase in demand has also led to an unintended effect: correspondingly large spikes in the amount of energy required.

Global electricity consumption from data centers, artificial intelligence and the cryptocurrency sector is expected to double from an estimated 460 terawatt-hours (TWh) in 2022 to more than 1,000 TWh in 2026, according to a research report from the International Energy Agency.

Researchers at the University of California, Riverside, published a study in April last year that found ChatGPT consumes 500 milliliters of water for every 10 to 50 prompts, depending on when and where the AI model is deployed. That equates to roughly the amount of water in a standard 16-ounce bottle.

As of August, there were more than 200 million people submitting questions on OpenAI’s popular chatbot ChatGPT every week, according to OpenAI. That’s double the 100 million weekly active users OpenAI reported last November.

Nuclear energy isn’t without its controversy. Many climate activists oppose such supplies, citing their hazardous environmental and safety risks, and the fact that they do not offer a genuine source of renewable power.

“Nuclear power is incredibly expensive, hazardous and slow to build,” the climate charity Greenpeace says on its website.

“It is often referred to as ‘clean’ energy because it doesn’t produce carbon dioxide or other greenhouse gases when electricity is generated but the reality is that it isn’t a plausible alternative to renewable energy sources.”

Proponents of nuclear energy, on the other hand, say that it offers a nearly carbon-free form of electricity and is more reliable than renewable sources like solar and wind.

“If it is built and securitized in the right way, I do think nuclear is the future,” Rosanne Kincaid-Smith, chief operating officer of Northern Data Group, a global data center provider, told CNBC at a tech conference in London last week.

“People are scared of nuclear because of the disasters we’ve had in the past. But what’s coming, I just don’t see traditional grids being the sustainable power that’s ongoing in the development of AI,” Kincaid-Smith added.

While Northern Data Group isn’t using nuclear energy — nor is it actively exploring plans to use nuclear as a power source for its AI data centers — the firm does want to “contribute to that conversation because it’s important for the wider ecosystem, the wider economy,” Kincaid-Smith told CNBC.

– CNBC’s Pippa Stevens contributed to this report