Google Goes Nuclear With Power Upgrade Agreement in Georgia

As the massive energy needs of the artificial intelligence (AI) build-out become increasingly apparent, AI companies and utilities alike are growing concerned.

Across the country, opposition to AI data centers has been rising, in part due to the strain these facilities are expected to place on local power grids and water systems.

Alphabet's (NASDAQ:GOOG) (NASDAQ:GOOGL) Google is the latest company to announce a deal to sweeten the pot for one of its proposed AI facilities by adding capacity to an electrical grid. And it's coming from a source that's been surprisingly popular with AI companies: nuclear power.

Here's what Alphabet is proposing and why nuclear power has become one of energy's hottest trends.

AI data centers have never been less popular

AI's reputation as a job-killer has always been a drag on its popularity. That existing stigma seems to have extended to AI data centers.

Residents of communities across the country have lobbied to have the projects canceled. In many cases, they've successfully pressured local politicians and zoning boards to rescind approvals, citing fears about the increased utility bills and pollution such projects would bring.

This kind of local opposition has already delayed or disrupted tens if not hundreds of billions of dollars' worth of AI infrastructure projects. That's prompted AI companies to add electricity generation infrastructure to their projects in hopes of making them more palatable to cost-conscious ratepayers.

But with the Trump Administration openly opposed to solar and wind energy development, an out-of-favor energy source is back in the mix: nuclear.

Upgrading instead of building

Nuclear power was originally adopted because it was cheap to produce – at least, after the enormous upfront costs of building the plant were covered.

Those economics changed as technological advances caused the price of natural gas, solar, and wind power to plunge. By 2018, in fact, a local utility in Iowa decided it was worth paying a $110 million fee to break its existing five-year contract with a local nuclear power plant and switch to cheaper wind and natural gas power.

But even though it's still cheaper to generate electricity using natural gas or wind turbines than nuclear reactors, U.S. electricity demand is outstripping supply. That's expected to make nuclear power competitive again in many areas.

The key is that it needs to be an existing nuclear power plant: the massive start-up cost of a new reactor remains prohibitive. That's why Google has inked an agreement with Georgia Power, the electric utility that covers most of Georgia, to pay for upgrades at two of its existing nuclear plants.

What Google has agreed to

Under the agreement, which has yet to be approved by the Georgia Public Service Commission, Google will pay for "uprates" at Georgia Power's Vogtle and Hatch nuclear plants.

Uprates are performance-improving upgrades that allow a nuclear plant's existing reactor core to operate at a higher temperature, producing more heat and thus more electricity. Uprates can include things like replacing turbines and introducing more efficient cooling systems.

Georgia Power expects the uprates to add about 96 megawatts of new capacity to the electricity grid, which could power tens of thousands of homes. Although the costs of the uprates haven't been specified, a press release from Georgia Power estimates that its customers will receive "approximately $900 million in projected benefits ... over the life of the units."

Google is betting big on nuclear

Naturally, given the current antipathy toward AI data centers, the press release doesn't mention the unspoken reason for the agreement: Google almost certainly intends to use some or all of the increased output to power its data centers in Georgia. Earlier this month, NextEra Energy secured a $1.9 billion government loan guarantee to restart a shuttered nuclear plant in Iowa to power a Google AI data center there.

With small modular reactor (SMR) companies like Oklo trying to develop data centers powered by on-site SMRs, nuclear power is becoming increasingly closely associated with AI. But it seems unlikely that nuclear power alone will be enough to power the hyperscalers' electricity-hungry AI ambitions, at least not in the near term.

John Bromels has positions in Alphabet, NextEra Energy, and Oklo. The Motley Fool has positions in and recommends Alphabet and NextEra Energy. The Motley Fool has a disclosure policy.