The need for new data centers is growing rapidly as tech companies speed up the creation of sophisticated AI models and the number of chatbot users increases dramatically. However, a significant obstacle to this growth is the enormous and frequently unsustainable electricity consumption of AI data centers.
Data center electrical supply has grown to be a major problem for businesses. Additionally, it has occasionally caused public outrage in the communities where these data centers are located. Google is looking for ways to lessen the burden on the US power system during times of high demand in response.
Google has struck two new utility agreements with Tennessee Valley Authority (TVA) and Indiana Michigan Power (I&M) to lower the power usage of AI data centers during peak hours, as the company announced in a blog post.
In order to control data center power use and lessen the strain on the US power grid when necessary, Google launched its demand response program in 2023.
Google says the demand response initiative "helps grid operators more effectively and efficiently manage power networks, reduces the need to build new transmission and power plants, and allows huge electricity loads like data centers to be integrated more quickly."
The earliest known example of lowering power consumption by reducing machine learning workloads is represented by Google's two most recent utility agreements.
"We are able to investigate opportunities at additional places because of our successful demonstration with Omaha Public Power District (OPPD), where we decreased the power demand associated with ML workloads during three grid incidents last year." Google was added.
By 2030, the energy consumption from AI data centers may treble, according to the International Energy Agency (IEA), and by the end of this decade, the technology could need almost as much electricity as Japan currently uses.
To fulfill the demands of their AI data centers, tech corporations are actively pursuing every feasible energy source, from increasing their investments in renewables to purchasing and constructing nuclear power plants.