Carbon footprint
Read the greenhouse gas emissions AWS, Azure and Google Cloud report for a workspace's connections, by month and by account, region or service, and export them as CSV.
Carbon footprint sits under Reporting in the product's sidebar. It shows the emissions each cloud provider itself reports for the workspace's AWS, Azure and Google Cloud connections, over the last twelve closed months: a total and its scopes, every month, and a breakdown by account, region or service. Every figure is the provider's own estimate under the provider's own methodology. Costfluent calculates no emissions, and it reads none for other providers, Kubernetes or AI APIs.

Before you start
- Any team role can view the page and export it. It is available on every plan; your plan's cost history limit applies to it as to every cost report.
- Each connection needs one extra step before its provider shares emissions with Costfluent; see What each provider needs. The cost data a connection already reads carries no emissions.
- Figures arrive once a month, on the 22nd. A connection set up today shows its figures after the next 22nd.
Read the report
- Open Carbon footprint under Reporting. The line above the figures names the months the report covers.
- Choose a method in the tabs: Location-based and Market-based for AWS and Google Cloud, Method not stated for Azure. Each tab shows only that method's figures.
- Read the method's total and its Scope 1, Scope 2 and Scope 3 split, then By month.
- Under Breakdown, choose Account, Region or Service in Break down by, and search the rows.
- Under Connections, read what the latest read of each connection found.
A method's figures read in one unit: kg CO2e while its total is under 1,000 kg, t CO2e from there. A row the provider left without an account, region or service reads "Not named by the provider".
Methods and scopes
A method is the way a provider counts the electricity behind your usage:
- Location-based counts electricity at the average emissions intensity of the grids it was drawn from.
- Market-based counts electricity at the intensity of what the provider bought, renewable energy contracts included.
- Method not stated is Azure's: Microsoft names no method for these figures, so they stand on their own.
Both methods count the same emissions, each in its own way, so Costfluent never adds two methods together. A workspace with AWS and Azure connections shows an AWS figure under each of the first two tabs and the Azure figure only under the third; no tab holds a grand total across methods.
The scopes are the provider's, as it reports them:
- Scope 1 is the provider's direct emissions, such as fuel burned in its data centers.
- Scope 2 is the electricity, heat and cooling the provider bought. This is the scope the method changes.
- Scope 3 is emissions up and down the provider's supply chain, such as building the hardware.
Each method's total holds all three scopes. Scope 1 is the same figure under both methods, and for Google Cloud so is scope 3. AWS states scope 3 per method.
What "Provider-reported, estimated" means
Every total carries Provider-reported, estimated. The provider estimates your share of its emissions from your usage, with its own model, and publishes the result; Costfluent reads it and converts its unit to kilograms of CO2 equivalent, nothing more. A provider can restate a month or publish under a new model version, so the same month can change between two reads. The CSV names the model version of every figure.
Each provider documents how it estimates:
- AWS: Carbon emissions calculation methodology
- Microsoft: Azure emissions calculation methodology
- Google Cloud: Carbon Footprint reporting methodology
This is not a compliance report. Costfluent does not calculate, verify, audit or certify the figures, and the report claims no reporting standard. Before you use a figure in a disclosure, read the provider's methodology for what it covers and leaves out.
Export the figures as CSV
- Open Carbon footprint.
- Select Export CSV. The file holds every figure of the months the report shows, for every connection and method, whichever tab is open.
One row per figure the provider reported, in these columns:
| Column | Holds |
|---|---|
month | The month, as 2026-09 |
provider | aws, azure or gcp |
connection | The connection's ID |
account | The AWS account, Azure subscription or Google Cloud project |
region | The region the provider names |
service | The service the provider names; for Azure, the resource type |
scope | scope 1, scope 2 or scope 3 |
method | location-based, market-based or not stated |
kg_co2e | The figure in kilograms of CO2 equivalent |
model_version | The provider's model version; empty for Azure, which states none |
basis | Always provider-reported estimate |
Adding up the rows of one method gives that method's total; adding rows across methods counts the same emissions twice. Each export is recorded in the audit log.
What each provider needs
| Provider | What to set up | Where |
|---|---|---|
| AWS | A second Data Export, of the CARBON_EMISSIONS table, into the same bucket and prefix as the cost export. The CloudFormation stack and the Terraform module create it; the role needs no new permission. | Connect AWS |
| Azure | The built-in role Carbon Optimization Reader for the service principal on each subscription, beside Cost Management Reader. | Connect Azure |
| Google Cloud | The Carbon Footprint export to BigQuery, into a dataset in the billing export's project, BigQuery Data Viewer on it for Costfluent's service account, and that dataset on the connection. | Connect GCP |
How it behaves
- Cadence. Costfluent reads every AWS, Azure and Google Cloud connection on the 22nd of each month, after all three have published the previous month: Google Cloud on the 15th, Azure by the 19th, AWS by the 21st. Each read takes the last three closed months again, because providers restate recent months. A connection that is in error or disabled on the 22nd is read on the next 22nd after it recovers, over the months it missed.
- First read. A connection's first read reaches back as far as the provider and your plan's history allow: to January 2022 for AWS, twelve months for Azure, and January 2021 for Google Cloud. The report shows the last twelve closed months of it.
- Plan history. When your plan's history starts inside the twelve months, the report starts there and says so: "Your plan's history starts in …, so this range starts there."
- Missing months. A month no connection of a method has published yet reads Unavailable, never zero. A month only some connections have published is marked "Not every connection has published this month", and its total holds only theirs.
- Connection status. Under Connections, each connection shows Collected, Not published yet, Not set up, Access refused or Not read yet, the months collected and when it was last read. A connection without figures carries the provider's sentence naming what is missing, such as the permission to grant and to whom.
- Azure subscriptions. One subscription without Carbon Optimization Reader refuses the whole month for its connection, because a month without it would read as complete. The sentence names the subscriptions.
- No connection. A workspace without an AWS, Azure or Google Cloud connection shows "No connection reports emissions" and a way to connect one.