Google partners with US utilities to reduce AI data centre power use

Google partners with US utilities to reduce AI data centre power use

The logo of Google is shown on a building in San Diego, California, U.S., October 9, 2024. Photograph: (Reuters)

Story highlights

AI technologies, particularly machine learning models, consume substantial amounts of energy, sometimes tens or hundreds of megawatts over extended periods. This intense demand has raised concerns as it strains power grids already facing supply constraints.

Google has signed agreements with two major US electric utilities, Indiana Michigan Power (I&M) and Tennessee Power Authority (TVA), to reduce its AI data centre power consumption during periods of high energy demand. This partnership is part of Google’s broader strategy to address sustainability concerns while continuing its aggressive expansion in AI technologies, which are known for their significant energy consumption. Under the terms of the agreements, Google will scale back its machine learning workloads at data centres when requested by the utilities, helping to free up capacity on the grid. This approach is known as “demand response”, a strategy typically used by energy-intensive industries like manufacturing or cryptocurrency mining. In exchange, businesses involved in such programmes often receive incentives like reduced electricity costs or payments.

Managing AI’s growing energy demands

AI technologies, particularly machine learning models, consume substantial amounts of energy, sometimes tens or hundreds of megawatts over extended periods. This intense demand has raised concerns as it strains power grids already facing supply constraints. With data centres accounting for a significant portion of Google’s energy use, the company’s commitment to energy flexibility could play a key role in stabilising grid capacity during peak periods.

Add WION as a Preferred Source

The collaboration with I&M and TVA marks the first formal demand-response agreements Google has entered into specifically for its AI workloads. Google aims to integrate flexibility into its energy usage, enabling faster deployment of new data centres while reducing the need for new power plants and transmission infrastructure. “This will help grid operators manage power use more efficiently,” the company said in a blog post.

Balancing sustainability with growth

While Google’s approach to curtailing non-essential AI workloads during high demand offers a solution for grid stability, the company is also pursuing alternative energy sources such as wind, solar, and nuclear power to further reduce its environmental footprint. Google’s sustainability efforts also include plans for small modular reactors, underscoring the company’s drive to find innovative ways to power its data centres sustainably.

Trending Stories

These partnerships come as Google continues to tackle the challenge of balancing rapid AI growth with environmental responsibility. Despite reducing its data centre energy emissions by 12 per cent in 2024, Google acknowledges that much more needs to be done to curb the carbon footprint of its expanding AI infrastructure.

(With inputs from agencies)