Software Carbon Intensity (SCI) Score
environmental
governance
systemic
The Software Carbon Intensity (SCI) Score, defined by the Green Software Foundation, measures the carbon emissions associated with a software system per functional unit (FU). The SCI has been standardized as an ISO standard, making it the internationally recognized reference for quantifying and comparing the carbon efficiency of software systems.
More details and the official specification are available from the Green Software Foundation.
It provides a comparable and transparent metric for understanding the environmental impact of software components across versions, architectures, or deployment environments. Because SCI normalizes emissions by a functional unit, it allows teams to track carbon efficiency independently of workload volume or user growth.
The SCI Score integrates three core elements:
- E — the energy consumed by the software
- I — the grid carbon intensity of the electricity used
- M — the embedded (manufacturing) emissions of hardware amortized across its lifetime
Software with a lower SCI Score is more carbon‑efficient. Tracking SCI across releases enables sustainable architectural decision‑making, supports carbon‑aware prioritization, and makes environmental trade‑offs visible.
- Category: Code, Build & CI/CD
- Measurement Frequency: per release
- Responsibility: Engineering Team (measurement), Product Lead (goal‑setting), Green Agile Coach (enablement)
Impact
The SCI Score helps teams understand how design choices — such as compute intensity, hosting region, model selection, data handling, or caching — affect the ecological footprint of their software. A decreasing SCI Score indicates that the system is becoming more carbon‑efficient.
By making emissions quantifiable, the metric supports sustainable architecture, carbon‑aware engineering, and responsible scaling. It also enables meaningful comparisons: between services, between releases, and between deployment regions. Because SCI incorporates both operational and embodied emissions, it encourages long‑term thinking and avoids shifting impact elsewhere in the system.
Embedded Emissions (M)
Embedded emissions describe the carbon footprint of the hardware used to run the software, including manufacturing, transport, and end‑of‑life processing. These emissions occur before the software is executed, but must be considered to avoid underestimating total impact.
In the SCI context, embedded emissions are typically amortized over the hardware’s lifetime and attributed proportionally to workloads (e.g. by CPU hours or capacity share). Even rough estimates help teams avoid shifting emissions from software efficiency gains to excessive or short‑lived hardware usage.
Calculation
The SCI Score follows the official formula defined by the Green Software Foundation:
\[\text{SCI} = \frac{(E \times I) + M}{R}\]where:
- E = energy consumed by the software
- I = grid carbon intensity (gCO₂e per kWh) of the region or provider
- M = embedded emissions of hardware (amortized)
- R = functional unit (e.g., per request, per job, per session)
A lower SCI is better, and teams aim for a decreasing trend across releases.
Example
Consider a backend service that processes API requests. During one release cycle, the following values are estimated:
- Energy consumption (E): 1 200 kWh
- Grid carbon intensity (I): 180 gCO₂e/kWh
- Embedded emissions (M): 60 000 gCO₂e (amortized share of server manufacturing emissions for this release)
- Functional unit (R): 40 000 API requests
Using the SCI formula with the measured values inserted directly:
\[\text{SCI} = \frac{(1\,200 \,\text{kWh} \times 180 \,\text{gCO₂e/kWh}) + 60\,000 \,\text{gCO₂e}} {40\,000} = 6.9 \,\text{gCO₂e per request}\]This value represents the average carbon intensity per API request for the current release.
If the team later improves caching and reduces energy consumption to 900 kWh while keeping the same functional output, the SCI Score decreases accordingly — demonstrating how efficiency improvements directly reduce carbon intensity per functional unit.