A large data centre is one of the biggest customers a grid can have. That can strain the grid at busy times, but it also creates an opportunity. Not all computing is urgent: some jobs can run an hour later or in another place. If a data centre eases its demand when the grid is stretched, it can help keep the lights on for everyone else, and it may be able to connect sooner. Several trials are now testing how much flexibility is really available.
Three ways to be flexible
A July 2026 report from the IEA 4E Technology Collaboration Programme groups data centre flexibility into three strategies: enabling workload flexibility, utilising supporting infrastructure, and integration of additional flexibility assets. It finds useful potentials of data centre flexibility exist, but that deployment is limited by operational and economic barriers.
| Strategy | What it means in practice |
|---|---|
| Workload flexibility | Shifting computing that can wait to another hour or another site |
| Supporting infrastructure | Using the site's own batteries, generators or cooling to reduce draw from the grid for a while |
| Added flexibility assets | Installing extra storage or generation so the site can respond more |
The US Department of Energy's best practice guide, written by NREL, notes that a data centre can respond to grid requests by switching to backup power generation.
What is happening now
Google reported in August 2025 that it had agreements with Indiana Michigan Power and the Tennessee Valley Authority to deliver data center demand response by targeting machine learning workloads, shifting non-urgent compute tasks during specific periods when the grid is strained. The Electric Power Research Institute's DCFlex initiative, whose members include operators, utilities and grid operators, describes itself as testing how data centers can flex in real time to support grid reliability and faster interconnections, through nine active demonstration sites.
The other side: when data centres unsettle the grid
Flexibility has to be planned. The companion article on power cuts describes how many data centres switching to backup at once in Virginia removed about 1,500 MW from the grid in July 2024. Planned, predictable flexibility helps the grid; sudden, uncoordinated changes can harm it.
What is not certain
- How much computing can really wait varies by customer and service; much AI work, such as answering users, cannot.
- Running backup generators for grid support has costs and local emission limits.
- The trials are recent, and their results were not yet fully published in October 2026.
Sources and method
- IEA 4E Technology Collaboration Programme, Electronic Devices and Networks Annex, Data Centres and Flexibility, 1 July 2026. https://www.iea-4e.org/edna/publications/data-centres-and-flexibility/
- Van Geet and Sickinger, NREL, for the US Department of Energy Federal Energy Management Program. Best Practices Guide for Energy-Efficient Data Center Design, revised July 2024. https://www.energy.gov/sites/default/files/2024-07/best-practice-guide-data-center-design_0.pdf
- Google, How we're making data centers more flexible to benefit power grids, 4 August 2025. https://blog.google/inside-google/infrastructure/how-were-making-data-centers-more-flexible-to-benefit-power-grids/
- Electric Power Research Institute, DCFlex. Read 29 October 2026. https://dcflex.epri.com
- North American Electric Reliability Corporation, Incident Review: Considering Simultaneous Voltage-Sensitive Load Reduction, 8 January 2025. https://nerc.com/pa/rrm/ea/Documents/Incident_Review_Large_Load_Loss.pdf
Method: each link was opened on 29 October 2026 and every figure was taken from the page cited beside it. No figure was estimated.
Read next
This article provides general information, not investment advice or an offer of securities.