×

Welcome to National Journal!

Enjoy this premium "unlocked" content until November 17, 2024.

Continue

AI's expansion could depend on an outdated energy source

Nuclear power has been on the decline in the U.S., but big tech and Congress see it as an answer to the growing energy demands of artificial intelligence.

Cooling ponds alongside Dominion Energy's North Anna Power Station along the shores of Lake Anna in Mineral, Virginia. (AP Photo/Steve Helber, File)
None
Nov. 4, 2024, 4:56 p.m.

With the massive expansion of artificial intelligence over the past two years, tech giants are turning to an often overlooked and slowly receding energy source—nuclear power—to feed the technology’s escalating appetite for electricity.

Most of the power demands for artificial intelligence are a result of the increasing strain from data centers, where massive amounts of computers and servers work to store the enormous amounts of data required to train AI programs and the computing power required to process AI demands.

These data centers already use a large amount of power. In 2022, data centers globally consumed an estimated 460 terawatt-hours, according to the International Energy Agency, or just under the annual energy use of Chile. The agency estimates that due to increased AI demand, the data centers could use 1,000 terawatt-hours of energy in 2026, which is roughly the equivalent of Japan’s total power consumption.

Stepping into the breach, big tech and Congress are turning to nuclear power, which has been on the decline for decades. Nuclear energy provides a way for these data centers to be run free of carbon emissions while meeting the rising electricity demand, advocates say.

The ability to provide power to data centers is shaping up to be a key factor in determining which nation wins the battle for AI supremacy.

In August, an Irish city council refused to grant planning permission to Google to build a new data center in south Dublin, largely because the Irish electrical grid simply can’t support more data centers. Industry insiders don’t expect Ireland to build a new data center until around 2030, possibly causing the country to miss out on a key part of the AI boom.

While Ireland grapples with data centers, Malaysia is cashing in by developing a number of data centers for both American and Chinese tech companies.

But nowhere in the world is better positioned for the AI boom than Virginia, which has more data centers—more than 400—than anywhere else in the world. Those data centers account for roughly 25 percent of Virginia’s total power consumption, according to the Electric Power Research Institute, and the state is starting to run out of energy capacity to build new ones.

The Virginia power company Dominion Energy said a lack of energy production and basic infrastructure will increase the time it typically takes a data center to connect to the grid: from between three and four years to between four and seven years, Bloomberg reported in August.

So far, three major tech companies—Microsoft, Google, and Amazon—have recently announced large investments in operators and developers of nuclear power plants in the wake of artificial intelligence's speedy growth.

While Amazon and Google chose to focus on small modular reactors, Microsoft took a different route. The company announced funding to revive the now-mothballed Three Mile Island nuclear power plant near Harrisburg, Pennsylvania, the same power plant that led America’s trust away from nuclear power.

Nuclear has been the subject of a massive, bipartisan push on Capitol Hill over the last year. The ADVANCE Act, coupled with the Price Anderson Act, which provides insurance in the event of a severe nuclear accident, were two major wins for the industry that came out of Washington.

But the support for nuclear power did not stop in Washington.

At last year’s Conference of the Parties of the United Nations Climate Change Conference, the U.S. and more than 20 countries pledged to triple nuclear energy capacity by 2050.

Jeremy Harrell, the CEO of ClearPath, a conservative nonprofit that advocates for market solutions to energy needs, explained why big tech companies decided to jump in on nuclear energy.

“They're competing in their data and AI businesses, and their data and AI businesses are contingent on the ability of being able to get electricity to grow fast, and so they're on a race,” Harrell told National Journal.

American trust in nuclear power hasn’t been the same since Three Mile Island's Unit 2 reactor suffered a partial meltdown in 1979, in what remains the nation's worst commercial industry accident.

“As we build more and more nuclear, that's only going to grow and change and continue to grow the national support totally, and the tech companies can play a big role in that,” Harrell said.

America’s nuclear industry has been on a slow but steady decline over the past few decades. The energy source accounts for a little more than 18 percent of total electric output, down slightly from about 20 percent only a few years ago, according to the Energy Information Administration.

Only three reactors have come online since 2010, whereas 13 commercial reactors have shuttered since 2013.

"Data centers, electric vehicles, and AI are driving a new surge in electricity demand after years of stagnant growth,” Senate Environment and Public Works Committee Chairman Tom Carper told National Journal. "The more we can meet that demand with clean electricity—including from new and existing nuclear power plants—the more momentum we will have in cutting greenhouse-gas emissions.”

Welcome to National Journal!

Enjoy this featured content until November 17, 2024. Interested in exploring more
content and tools available to members and subscribers?

×
×

Welcome to National Journal!

You are currently accessing National Journal from IP access. Please login to access this feature. If you have any questions, please contact your Dedicated Advisor.

Login