Tech Giants Turn to Nuclear Energy to Power AI Infrastructure Amid Rising Demand
Tech giants are increasingly turning to nuclear energy to power the energy-intensive data centers required for training and operating large-scale AI models driving generative AI applications.
Microsoft and Google have secured agreements with U.S. nuclear power suppliers to bring additional energy capacity online for their data centers. This week, Google announced a deal to purchase power from Kairos Power, a developer of small modular reactors, to support its AI operations.
“The grid needs clean, reliable sources of energy to support the growth of these technologies,” said Michael Terrell, Google’s senior director for energy and climate. “Nuclear energy can play an important role in meeting our demand sustainably and providing round-the-clock energy.” Google expects its first reactor from Kairos Power to be operational by 2030, with additional reactors planned through 2035.
Microsoft is also pursuing nuclear solutions. Last month, the company reached an agreement with Constellation Energy to restart a dormant reactor at the Three Mile Island plant in Pennsylvania, which has been offline for five years. Notably, this facility was the site of the most serious U.S. nuclear incident in 1979, caused by a loss of coolant and overheating of the reactor.
The push toward nuclear energy comes as tech firms seek sustainable energy sources to power the growing number of data centers supporting cloud computing and AI services. Companies like Amazon, Microsoft, and Google, which provide rented servers equipped with powerful GPUs, have benefited from the rising demand for AI platforms, including tools like ChatGPT. However, this surge has resulted in steep increases in energy consumption.
According to the International Energy Agency, global electricity use by data centers, AI, and cryptocurrency industries is projected to double from 460 terawatt-hours (TWh) in 2022 to over 1,000 TWh by 2026. A 2022 study by the University of California, Riverside, found that ChatGPT consumes 500 milliliters of water for every 10 to 50 prompts, comparable to the volume of a standard 16-ounce water bottle. With more than 200 million people using ChatGPT weekly as of August, energy and resource consumption continues to climb.
While nuclear energy offers a low-carbon power source, it is not without opposition. Environmental activists warn of the high costs, safety risks, and lengthy construction times associated with nuclear plants. Greenpeace, for instance, acknowledges that nuclear does not emit carbon during energy generation but argues it cannot replace renewable sources like wind and solar.
Proponents, however, emphasize nuclear power’s reliability compared to renewable sources. “If developed and managed properly, nuclear could be the future,” said Rosanne Kincaid-Smith, COO of Northern Data Group, a global data center provider. Although her company is not currently pursuing nuclear energy, Kincaid-Smith stressed the importance of engaging in discussions about nuclear power’s role in shaping sustainable infrastructure for AI. “Traditional grids won’t be enough to sustain future AI development,” she added.
news via inbox
Get the latest news, expert insights, success stories and updates straight to your inbox