Greenpixie and Microsoft calculate the impact of your cloud usage in different ways. Their methodologies differ in approach, granularity, timeliness and data sources, so which one to use depends on the outcome you need and the use case in front of you.
How does Azure carbon optimization work?
Azure carbon optimization is a free tool in the Microsoft Azure portal that reports the carbon emissions of your Azure resources. It starts from the total emissions of an Azure datacenter region and divides that total among customers, so the figure is a share of Microsoft's own data centre emissions and not a measurement of your resources.
The allocation works in three steps:
- Microsoft totals a data centre region's emissions, covering Scope 1, market-based Scope 2 and Scope 3 categories 1, 2, 4, 5, 9 and 12.
- They then work out each customer's share of that region's usage, where usage is defined in Microsoft's own white paper as "the normalized cost metric associated with infrastructure as a service (IaaS)", so spend, normalised to remove discounts, with a proportionate share of overhead server capacity added on.
- Then they multiply the region's emissions by that share to give each customer its figure.
Microsoft's own FAQ says usage for emissions "might not match your Microsoft usage for billing purposes", and it does not say how the two differ. No usage factor values are published. Coverage is described as "all Azure resource types, based on billing and usage"; no service list is published, and Microsoft does not say whether Azure OpenAI, AI Foundry models, GPU virtual machines, Databricks, Fabric or Marketplace software are included.
Scope 2 appears under the label "Scope 2 Market" and no other label. The methodology page names neither the market-based nor the location-based method, and it refers to grid emission factors and carbon factors without listing them. No location-based figure exists for customers. Hardware equipment lifetime defaults to six years, and the methodology does not currently include the data centre facility.
The customer receives emissions for Scope 1, Scope 2 Market and Scope 3, by subscription, resource group, resource, resource type and location. Emissions arrive monthly, with the previous month available by day 19 of the current month. No kWh or water figure appears in any customer view. Microsoft names carbon optimization as the recommended tool.
What is Azure carbon optimization good for?
For annual market-based reporting, it gives you carbon emissions for Scope 1, Scope 2 Market and Scope 3 across your Azure estate at no cost. The API provides up to 12 months of data, and Microsoft encourages you to export it regularly to keep a longer history.
What can't Azure carbon optimization be used for?
Below are some common use cases our customers have told us they struggle to accomplish with the native tool.
| Use case | Why Azure carbon optimization falls short | With Greenpixie |
|---|---|---|
| Confirming that rightsizing reduced emissions | Because emissions data is updated monthly, changes to a resource configuration "appear in the emissions data of the following month". | Per resource, per day |
| Checking whether AI, GPU and Marketplace services are included | Microsoft does not say whether these services are included or how they would be reported. | Every GPU instance family, plus per-token AI data |
| Measuring electricity per workload | Microsoft shows no kWh for a workload, and the API has no field for energy. | kWh on every line item |
| Reporting location-based Scope 2 | A disclosure that needs both the market-based and the location-based method needs another source. | Location-based Scope 2 |
| Reporting water use | No Azure customer view shows water. | Two water columns per line item |
| Reconciling emissions to your Azure bill | Microsoft defines carbon intensity as your allocation of its emissions divided by your cloud usage hours. The usage in that calculation is the cost-derived measure described in the section on how Azure carbon optimization works. | Calculated per SKU from the billing export, line by line |
| Reporting emissions per customer | Azure carbon optimization breaks emissions down by subscription, resource group, resource and location, but not by tag, and the values are monthly. A software vendor who needs a per-customer or per-tenant figure has to map resources to customers themselves, from rounded monthly totals, with no usage quantity or energy figure to allocate on. | By tag or account |
| Comparing Azure with other cloud providers | The FinOps Foundation notes that providers "report using different methods". An Azure figure next to an AWS figure compares two methodologies, so the figures are not like for like and the comparison is not useful. | One method across AWS, Azure and Google Cloud |
| Planning a workload for lower emissions | Carbon optimization reports past emissions and takes its recommendations from Azure Advisor, which shows the carbon and cost savings of changes such as shutting down or rightsizing idle virtual machines, based on their resource utilization. The page does not explain how the carbon savings are calculated, and the data is monthly, with nothing by hour. | Hourly grid intensity per region |
| Choosing an architecture for lower impact | Because emissions are allocated on a normalised cost metric, a cheaper architecture looks lower carbon without using less electricity. | Model a change from the rate card |
| Choosing between AI models | Choosing a model on its footprint needs energy per token by model and inference region, which Azure carbon optimization does not provide. | Energy per token for 500+ models |
What Greenpixie adds
Greenpixie adds energy, carbon and water columns to every line of the Azure usage detail or cost export that your FinOps team already works with, and leaves annual market-based reporting to Azure carbon optimization.
For AI, every GPU instance family is included, and the AI token methodology gives energy, carbon and water per token, across scopes 1, 2 and 3, for over 500 open and closed models in varying configurations. The methodology is built on hardware benchmarking and published as a white paper.
Finding which workloads emit the most
Greenpixie calculates at the individual SKU level, at the same granularity as the billing export by day and quantity, joined on the SKU code. It typically covers 99% to 100% of relevant SKUs by spend, and you can allocate carbon to a team, product or customer in the same way as cost.
Reading electricity in kWh per line item
Each line item carries electricity in kWh, alongside usage emissions and embodied emissions in tCO2e.
Reporting location-based emissions and water use
Greenpixie reports Scope 2 on the location-based method, and each line item has two water columns, one for cooling water and one for the water used to generate the electricity.
Scheduling batch jobs and choosing regions by hour
Greenpixie applies grid carbon intensity at hourly resolution for the grid area supplying each region, so the same virtual machine gets a different carbon figure at 2pm and 2am. That makes time-shifting and region selection measurable.
Modelling a change before making it
The Greenpixie API serves rate cards with energy, carbon and water per unit for each SKU. You join them to your own billing data on the SKU code at whatever refresh rate you want, up to every 6 to 12 hours, and swapping the SKU models a change before you make it.
Seeing the data inside FinOps tools
Greenpixie's energy, carbon and water columns are generally available inside Flexera, CloudHealth, Cloudability, and Umbrella.
How Azure carbon optimization and Greenpixie fit together
| Question | Azure carbon optimization | Greenpixie |
|---|---|---|
| What was our Azure footprint last year, for the annual report? | Yes, market-based Scope 2 | Supporting detail |
| Which subscriptions and regions carry the most carbon? | Yes, monthly | Yes, down to the resource |
| How many kWh did this workload use? | No | Yes, per line item |
| What is our location-based Scope 2? | Microsoft does not show one | Yes |
| Did last month's rightsizing reduce emissions? | Monthly, with changes appearing the following month | Yes, per resource, per day |
| When should this batch job run? | No | Yes, hourly intensity |
| What did our Azure OpenAI and GPU virtual machines emit? | Microsoft does not say whether they are included | Per instance, and per model and token type |
| How much water did this workload use? | No | Yes |
| Can we report it to a customer, per tenant? | No | Yes, by tag or account |
FAQs
Which Azure services does carbon optimization cover?
Microsoft publishes no service list, and the section on how Azure carbon optimization works names the services that Microsoft does not address.
Does Azure carbon optimization include Scope 3?
Yes, for categories 1, 2, 4, 5, 9 and 12, and Microsoft's cloud hardware emissions methodology white paper sets out how hardware emissions are calculated.
How different are Greenpixie's and Microsoft's numbers for the same subscription and period?
Microsoft allocates a share of its emissions to you by relative usage in a region. Greenpixie estimates what each resource drew from the grid, day by day, so the figures can differ a lot depending on what you run. It provides an audit trail for every number, by resource ID, usage type and day.
Which carbon factors does Azure carbon optimization use?
The methodology page names grid emission factors and carbon factors for cloud infrastructure, without listing the values.
Why does my data show the same figure every month?
One practitioner who tested the export found that it "contains rounded values, not raw ones", so a resource can show the same 0.1 kgCO2e every month. The portal table shows the last two months.
Learn more
To learn more about Greenpixie's methodology and work with the data, take the Greenpixie Data Explorer course at the Greenpixie Academy.
For the other clouds, read AWS Sustainability vs Greenpixie and Google Cloud Carbon Footprint vs Greenpixie. For an open-source option, read Cloud Carbon Footprint and its alternatives.
Measure your cloud emissions
See the energy, carbon and water behind your cloud bill
GPX Cloud adds energy, carbon and water data to AWS, Azure and Google Cloud billing data, line by line.
Sources
All statements about Azure carbon optimization were checked against the sources below and are accurate as of Nov 10, 2026.
Microsoft
- Carbon optimization overview. Microsoft Learn
- View emissions. Microsoft Learn
- Emissions terminology. Microsoft Learn
- Reduce emissions. Microsoft Learn
- Troubleshooting. Microsoft Learn
- Export emissions data with the API. Microsoft Learn
- Azure emissions calculation methodology. Microsoft Learn
- A new approach for Scope 3 emissions transparency. Microsoft white paper, 2021
- Generally Available: Carbon optimization in Azure. Azure Updates, mirror, May 2025
- Retirement: Emissions Impact Dashboard for Azure. Azure Updates, mirror
- Cloud hardware emissions methodology. Microsoft white paper, January 2026
Third parties
Greenpixie
- Methodology overview. Greenpixie documentation
- AI token methodology. Greenpixie documentation
- Greenpixie's AI Token Methodology: assessing the energy, water and CO2-eq impact of AI tokens. arXiv white paper
- Grid and regions methodology. Greenpixie documentation
- CUR enrichment. Greenpixie documentation
- Greenpixie API. Greenpixie documentation
- GPX Cloud. Greenpixie product page
- Flexera bolsters FinOps offering with Spot Eco, Ocean and Greenpixie partnership. Flexera press release, May 2025
- Greenpixie Data Explorer course. Greenpixie Academy

