Google has signed agreements with two power companies to implement 'demand response' to reduce AI workloads in data centers during peak demand periods or when power supply is reduced due to bad weather.
Google has announced that it will implement a feature called ' demand response ' that will reduce AI workloads on its data centers during times of peak demand or when power supply is reduced due to bad weather.
How we're making data centers more flexible to benefit power grids
Google agrees to pause AI workloads when power demand spikes • The Register
https://www.theregister.com/2025/08/04/google_ai_datacenter_grid/
Google explains that meeting the energy needs of AI efficiently and reliably presents a unique opportunity to modernize the entire energy system. To that end, the company has been working to implement flexible demand response capabilities in its data centers, which will enable it to shift or reduce electricity demand at certain times of the year.
This flexible ability to shift electricity demand is called demand response , and it can provide several benefits as demand for electricity continues to grow, especially in the United States and around the world. It not only allows large power loads like data centers to be interconnected more quickly, reducing the need to build new transmission and power plants, but it also helps grid operators manage the grid more effectively and efficiently.
Google has signed two new utility contracts with Indiana Michigan Power (I&M) and the Tennessee Valley Authority (TVA) to offer data center demand response targeted at machine learning workloads for the first time. This builds on a successful 2024 trial with Omaha Public Power Department (OPPD), which successfully reduced electricity demand for machine learning workloads.
'We are pleased to partner with Google to enable demand response at their new data center in Fort Wayne, Indiana,' said Steve Baker, president and COO of I&M. 'When adding large new loads to the system, it is essential to work with our customers to effectively manage the necessary generation and transmission resources. Google's ability to leverage load flexibility as part of their unplanned load response strategy will be an invaluable tool in meeting our customers' future energy needs.'
Google pioneered the demand response feature in its data centers, which offloads less time-critical computing tasks, such as processing YouTube videos, during certain times of high load on the grid. Through ongoing partnerships with Belgium's Centrica Energy, grid operator Elia, and Taiwan's Taiwan Power Company, the company is leveraging this capability to help grid operators maintain reliable power during their highest demand times of the year.
'As AI adoption accelerates, Google sees significant opportunities to expand its demand response toolkit, develop specialized capabilities for machine learning workloads, and leverage them to manage large-scale new energy loads. By incorporating load flexibility into overall energy planning, even regions with generation and transmission constraints can accommodate AI-driven growth.'
Data center demand flexibility is still in its infancy and is only available in select locations. High reliability is essential for cloud customers in key industries like healthcare, as well as services like search and maps, limiting the flexibility of any given data center.
The deployment of machine learning workloads is an important step toward enabling demand flexibility at a larger scale, delivering grid reliability and cost savings benefits in the areas where these capabilities are deployed. By developing long-term resource plans with utility partners like I&M and TVA, we can build flexibility into future grid developments in tandem with the deployment of Google's data center infrastructure.
in Note, Posted by logu_ii