# What is the Software Carbon Intensity (SCI) score and how do I calculate it?

Canonical: https://senseible.earth/answers/software-carbon-intensity-sci-explained

The SCI score is a rate: the carbon emitted per unit of work your software does, such as per API call or per active user. ISO/IEC 21031:2024 sets the formula: energy used times grid carbon intensity, plus a share of the hardware's embodied emissions, divided by your chosen functional unit.

## The formula, term by term

SCI = ((E x I) + M) per R.

- E: energy the software used over the period, in kWh, for the servers, storage and network you include.
- I: carbon intensity of that electricity, in grams per kWh, for the grid region where it ran.
- M: embodied emissions of the hardware, shared out by time and by the part of the machine you reserved.
- R: the functional unit, the thing you divide by, such as one API call, one user, one transaction or one GB processed.

## How to share out embodied emissions

M = total embodied emissions of the device x (hours you reserved it / expected life in hours) x (resources you reserved / total resources). On a shared cloud machine the last part is your vCPUs out of the host's vCPUs.

Few cloud providers publish embodied figures per instance, so most teams use a manufacturer's product footprint or a published estimate and label it as an assumption.

## What the score is not

- It is not your company footprint. A company inventory reports total tonnes by scope; the SCI is a per-unit rate for one software system.
- Offsets and renewable energy certificates cannot lower it. The specification excludes all market-based measures, so I must be the location-based intensity of the grid you draw from, and only cuts in energy, hardware or that grid intensity move the score.
- It is not comparable across products unless the boundary and functional unit match. Publish both alongside the score.

## Choosing the functional unit

Pick the unit your product scales with and that a customer understands. An API business uses calls, a collaboration tool uses active users, a data pipeline uses GB processed. Keep the same unit over time so the trend means something.

## Worked example: one month of an API service in India

- Energy: one dedicated server averaging 1 kW for 720 hours = 720 kWh
- Operational: 720 x 0.710 kg CO2 per kWh (CEA, version 21.0) = 511.2 kg
- Embodied: assumed 1,500 kg CO2e for the server over a 4 year life (35,040 hours); 1,500 x 720 / 35,040 = 30.8 kg
- Total: 511.2 + 30.8 = 542.0 kg
- R: 2,000,000 API calls; 542,000 g / 2,000,000 = 0.271 g per call

The embodied figure is an illustrative assumption. The grid factor covers CO2 only; say so if you report the score as CO2e.

## What Senseible does here

Senseible's cloud and data centre calculator turns vCPU-hours, memory, storage and network transfer into kWh and emissions by hosting region, which gives the E x I part. For servers on your own premises, it reads the kWh from your electricity bill and applies 0.710 kg CO2 per kWh.

- [Cloud and data centre emissions calculator](https://senseible.earth/calculators/cloud-data-centre-emissions)
- [Cloud footprint from AWS, Azure and Google](https://senseible.earth/answers/cloud-carbon-footprint-aws-azure-google-india)
- [Colocation emissions with PUE](https://senseible.earth/answers/colocation-data-centre-emissions-with-pue)

Sources: [ISO/IEC 21031:2024, Software Carbon Intensity (SCI) specification, published by the Green Software Foundation as SCI v1.1.0 (iso.org)](https://www.iso.org), [CEA CO2 Baseline Database](https://cea.nic.in/cdm-co2-baseline-database/?lang=en)

Last reviewed 2026-10-18.