Search

By EPN Staff

Google announced a pair of new utility agreements this month that may provide a roadmap for data center projects looking to throttle their energy usage so they can hook into existing power grids faster and with less disruption for the overall grid.

The tech giant has agreed to scale back power usage when needed to free up space on the grid by dialing back machine learning functions that aren’t time-critical. These agreements, with Indiana Michigan Power and the Tennessee Power Authority, are some of the first of their kind as electric utilities and the data industry grapple with massive new power demands for artificial intelligence projects.

A well-known researcher and consultant in this space, Tyler Norris at Duke University, said the development “could begin to reshape the relationship between AI infrastructure and electricity planning.”

Why it matters

Predictions on how much more electricity society will need as artificial intelligence is layered into the economy vary, but everyone agrees: Demand will go up significantly.

That leaves important questions about how the costs of new generation and grid infrastructure are spread out between new AI customers and regular ratepayers.

Norris’ team at Duke University published research last year arguing that much of the coming AI demand can be handled by existing power systems, if those AI projects are willing to cut their electricity usage when demand spikes elsewhere.

Google’s new utility agreements, which follow a demonstration project with the Omaha Public Power District, “represent the first time we’re delivering data center demand response by targeting machine learning (ML) workloads,” Google said in its announcement.

Some Google functions will run 24/7, taking priority. But others, like processing a YouTube video, can be delayed when the grid is strained, the company said.

This demand flexibility will be incorporated at limited locations for now, but the company called the capability “a promising tool for managing large new energy loads and facilitating investment and growth.”

The bigger picture

The initiative is part of Google’s carbon-free energy push, but it isn’t just a power-saving measure. This sort of flexibility will likely help large-load users, like data centers, jump the line to connect faster to power grids otherwise concerned about providing that much juice to a new customer.

“It allows large electricity loads like data centers to be interconnected more quickly, helps reduce the need to build new transmission and power plants, and helps grid operators more effectively and efficiently manage power grids,” Google said in its announcement.

Norris, the Duke University researcher, said these contracts are the first he’s seen between a hyperscaler and a U.S. utility built around data center flexibility.

Norris said the contracts are “definitive and long-term in nature” and that they will “establish official precedent for definitive long-term contractual arrangements” between data centers, utilities and grid operators.


Subscribe to our newsletter: