1. The ISO/IEC 21031:2024 Software Carbon Intensity (SCI) Equation
Standardized by the Green Software Foundation and published as ISO/IEC 21031:2024, the Software Carbon Intensity (SCI) specification quantifies the rate of carbon emissions per functional unit ($R$) of an AI or cloud workload (such as per 1,000 inference tokens, per model training epoch, or per API call):
Where:
- $E$ = Operational energy consumed by the AI cluster in kilowatt-hours ($\text{kWh}$), scaling raw GPU/CPU/DRAM draw by the data center Power Usage Effectiveness (PUE):
- $I$ = Location-based or time-matched marginal grid carbon intensity ($\text{gCO}_2\text{e/kWh}$).
- $M$ = Embodied Hardware Emissions ($\text{gCO}_2\text{e}$) of GPU accelerators, host servers, and NVMe storage allocated over the workload execution window ($\Delta t$) relative to expected hardware lifespan ($L$) and resource share ($\text{RS}$):
- $R$ = The functional scaling unit (e.g.,
1,000 inference requests).
2. Reconciling SCI Scores vs. GHG Protocol Inventory Scopes
A critical distinction for corporate sustainability teams is that SCI is a software engineering rate metric (which intentionally excludes market-based EACs/PPAs so software engineers cannot optimize code by buying certificates), whereas GHG Protocol Scope 2 & Scope 3 Category 1 require corporate inventory accounting:
| Deployment Model | GHG Protocol Inventory Classification | Required Telemetry & Cloud Provider Data | Water Footprint Metric (WUE) |
|---|---|---|---|
| On-Premise / Colocation AI Cluster | Scope 2 (Purchased Electricity) + Scope 3 Cat 2 (Capital Goods GPU embodied LCA) | PDU / BMC telemetry + colocation provider PUE and meter statements | Onsite cooling tower WUE ($\text{Liters} / \text{kWh}_{\text{IT}}$) |
| Hyperscale Cloud IaaS / Managed LLM API | Scope 3 Category 1 (Purchased Goods & Services — Cloud Compute) | AWS Customer Carbon Footprint Tool, Google Cloud Carbon Footprint, or Azure Emissions Impact Dashboard | Direct cooling WUE + Indirect grid electricity hydro/thermoelectric water factor (EWIF) |
3. Data Center Water Usage Effectiveness (WUE) under ISO/IEC 30134-9
AI clusters operating at high rack power densities (40–100+ kW/rack) rely heavily on Adiabatic or Evaporative cooling towers alongside closed-loop liquid cooling. Per ISO/IEC 30134-9, Water Usage Effectiveness (WUE) measures direct site water withdrawal per unit of IT equipment energy:
Enterprise AI sustainability assessments should evaluate both direct cooling water consumption ($\text{WUE}_{\text{site}}$) and indirect off-site water evaporated during thermoelectric power generation.
Geographic & Regulatory Applicability (UK / EU / US / UAE): Data center energy and water telemetry is governed by the EU Energy Efficiency Directive (Recast Directive (EU) 2023/1791) for facilities $\ge 500\text{ kW}$, US EPA eGRID regional factors, UK DESNZ grid schedules, and UAE DEWA / EWEC data center electricity factors. Screen facility Scope 2 electricity loads in our Corporate GHG Screening Calculator per Methodology v1.0.0 and compare cloud-integrated platforms in the Carbon Accounting Software Directory.