Cloud Workload Carbon Calculator

Because they answer a different question. AWS Sustainability, Google Cloud Carbon Footprint and Azure carbon optimization measure what you already emitted, for their own cloud, a few weeks after the month ends. None of them tells you what a workload you have not deployed yet would emit, what it would cost in a cleaner region, or how two clouds compare. Once the workload is running, use the provider tool: it measures what this estimates. To compare their reports across clouds, use the Cloud Carbon Report Analyzer.

Among Google's 42 published regions, northamerica-northeast1 (Montréal) is the lowest at about 11 gCO₂e/kWh, and africa-south1 (Johannesburg) the highest at 641: about 58 times as much for the same work. AWS and Azure regions on the same grids share those figures. The grids of Quebec, France, Sweden, Switzerland and Norway are mostly hydro or nuclear; the coal- and gas-heavy grids of South Africa, India, Poland and the Gulf are at the other end. Country averages from Ember are lifecycle figures, so they are ranked only among themselves. The region finder ranks them across all three clouds with prices.

Because the figure belongs to the electricity grid, not to the cloud. Google publishes the grid intensity for each of its regions; AWS and Azure publish none. A data center in Northern Virginia draws from the same grid whoever owns it, so AWS us-east-1 and Azure East US take Google's Northern Virginia figure. Where no Google region shares the grid, the country's average from Ember is used if the country's grid is close to uniform, and otherwise the region shows no figure at all. Each region says which applies.

Less than the region. The provider affects data center efficiency (a PUE of 1.1 to 1.185 in the published coefficients), the assumed watts per vCPU and the copies each storage service keeps. For compute the cloud moves the estimate by up to about 15%; for object storage by up to 3 times, from the copies each service keeps. These differences come from the model, not from measured data centers. The grid moves it by a factor of up to 58.

They are two accounting methods. Grid carbon intensity is location-based: what the grid where the data center sits emits per kilowatt-hour. Carbon-free energy percentage is Google's hourly share of carbon-free electricity in a region, counting both the grid's own carbon-free supply and the clean energy Google buys there. Google's us-central1 has 88% carbon-free energy and a 432 gCO₂e/kWh grid (2025), and both are true. This calculator uses grid intensity, because it means the same thing whoever owns the building.

Only if you tick "Include making the hardware". It adds each instance type's share of its server's manufacturing emissions, spread over a four-year life, from Cloud Carbon Footprint's published coefficients. Provider reports include manufacturing emissions, so tick it to compare with them. Custom sizes and newer instance types without a published figure are left out, and the page says how many.

It is an estimate, shown as a range on purpose: Cloud Carbon Footprint publishes a minimum and maximum watts per vCPU because the true figure depends on the processor and how busy it is. The low end is idle and the high end full load: scenarios, not a confidence interval, and the actual figure can fall outside them. The comparison between regions is far more reliable than the absolute figure, because the uncertain terms are the same in both.