Google data centers shift their computations to cleaner times and locations
Google's carbon-aware computing platform uses forecasts from Electricity Maps to run compute-heavy tasks at times where the grid is cleaner.
The Challenge: Aligning compute demand with clean electricity supply
The fight against climate change demands transformation across every layer of technology, including the invisible infrastructure powering the modern internet. Google set an ambitious goal: operate all of its data centers on 24/7 carbon-free energy (CFE), matching electricity consumption with locally sourced clean energy every single hour of every day.
This goes far beyond annual renewable energy credits or offsets. It requires real-time alignment between when and where compute tasks run, and when and where clean electricity is actually available on the grid, from wind in Denmark, to hydro in Oregon, to solar in Chile.
Achieving this requires knowing, in advance, how the carbon intensity of each local grid will change over the coming hours. Without that foresight, there is no way to systematically prioritize clean energy at scale.
The Solution
Shifting flexible workloads with real-time carbon data
Google integrated Electricity Maps' hourly carbon intensity forecasts into its carbon-intelligent computing platform. Electricity Maps combines live grid data, weather modeling, and energy systems analysis to predict, with high granularity, how clean or carbon-intensive the electricity supply will be at each data center location throughout the day.
Google's platform pairs these carbon forecasts with its own internal power forecasts, which estimate how much compute capacity each data center will need over the same period. Together, these two inputs enable a powerful optimization: generating hour-by-hour guidelines for when to run energy-intensive workloads based on the expected carbon intensity of the local grid.
Flexible compute tasks, such as training AI models, running data analytics pipelines, or processing batch jobs, can be automatically scheduled for the cleanest hours. When the local grid is forecasted to be carbon-heavy, workloads can also be shifted geographically to data centers in regions where clean energy is most available at that moment.
How It Work: From forecast to workload scheduling
Carbon intensity forecasting Electricity Maps delivers hourly predictions of grid carbon intensity for each data center location, drawing on live grid data, weather patterns, and energy modeling.
Internal power forecasting Google's systems generate a complementary forecast of each data center's expected power demand and compute needs across the same time window.
Combined optimization Both forecasts are combined to produce hour-by-hour scheduling guidelines, aligning compute tasks with the lowest-carbon electricity periods available.
Global workload shifting When local grids are carbon-intensive, flexible tasks are shifted to data centers in regions with cleaner energy, optimizing across both time and space.
Automated, reliability-preserving execution Guidelines are passed to workload schedulers and executed automatically, without compromising service performance, uptime, or user experience.
Impact: A working model for sustainable digital infrastructure
This system is already deployed across several of Google's largest data centers and is being scaled globally. What makes it significant is not only its environmental impact, but its proof of concept: sustainability can be built into infrastructure operations without sacrificing performance or reliability.
As grids continue to decarbonize and renewable variability increases, carbon-aware computing becomes an increasingly essential tool, one that ensures organizations make the most of every green electron already on the grid, without waiting for future technologies or hypothetical solutions.
Lower emissions per compute cycle Workloads timed to cleaner electricity windows reduce CO₂ per unit of compute, directly cutting Google's operational carbon footprint.
Better use of existing clean energy No need to build additional infrastructure. The system maximizes the clean energy already flowing through the grid.
Progress toward 24/7 CFE Hour-by-hour alignment of demand and clean supply advances Google's most ambitious energy goal without compromising operations.
A replicable model for the industry This collaboration demonstrates how real-time data and intelligent automation can make sustainability a built-in feature of digital infrastructure, not an afterthought.
