Data centers are learning to move flexible workloads through time, following cleaner and cheaper electricity across the grid.
When compute follows the weather
Google’s carbon-intelligent computing platform began by shifting non-urgent work to times when a data center’s local grid was cleaner. The company described aligning forecasts of hourly carbon intensity with forecasts of compute demand.
The insight is simple: not every computation has the same deadline. Training jobs, batch inference, indexing, and media processing often have temporal flexibility. A scheduler can spend that flexibility when lower-carbon electricity is more available.
Time, location, and pressure on the grid
Google later extended the idea across locations, reserving capacity where and when electricity was cleaner. Its current cloud architecture guidance describes temporal shifting as a practical sustainability technique and notes that carbon-aware scheduling depends on timely information about regional energy mixes.
Carbon is not the only signal. Similar scheduling can respond to price, grid congestion, water stress, and contractual power limits. The difficult part is deciding which workloads can move without making the product worse for people.
The operational metric to watch
A credible claim should report more than annual renewable-energy purchases. Look for hourly matching, the share of load that can move, the geographic boundaries of the system, and whether efficiency gains are being outpaced by total demand.
Read it for yourself.
Every source used in this dispatch is linked directly. Open the original material, inspect the claim, and draw your own conclusion.
- 01Primary source · April 22, 2020Our data centers now work harder when the sun shines and wind blowsGoogle
- 02Primary source · Reviewed January 28, 2026Optimize resource usage for sustainabilityGoogle Cloud
Explainer based on Google’s disclosed production approach and current cloud guidance. It does not estimate the net emissions of a specific AI service.