The GHG Protocol — the global standard for corporate greenhouse gas accounting — requires software and technology companies to measure emissions across three scopes, but the specific emission sources, calculation methodologies, and material categories differ significantly from traditional industrial companies. In 2026, with the EU CSRD and SEC climate disclosure rules in effect, accurate Scope 1/2/3 measurement is a financial reporting requirement, not a voluntary initiative. This guide covers what software companies must measure, how to calculate it, and how to build the data infrastructure to do it reliably.
Scope 1, 2, 3 for Software Companies
| Scope | Definition | Material Sources for Tech Companies |
|---|---|---|
| Scope 1 | Direct emissions from sources owned/controlled | Office heating (gas boilers), company vehicle fleet, on-site generators, fire suppression (HFCs) |
| Scope 2 | Indirect emissions from purchased energy | Electricity for offices, owned data centres. Market-based (RECs/PPAs) vs location-based methodologies |
| Scope 3 | All other value chain emissions | Purchased cloud/co-lo compute (Cat 1), business travel (Cat 6), employee commuting (Cat 7), hardware supply chain (Cat 1), customer product use (Cat 11), sold software's AI inference (Cat 11) |
Why Scope 3 Is the Hard Part for Tech Companies
For most software companies, Scope 1 and 2 are relatively small — office electricity and some travel. Scope 3 is where 70–95% of total emissions sit. Category 1 (purchased goods and services) includes the carbon footprint of all your cloud computing — AWS, GCP, Azure are your largest supply chain emissions source. Category 11 (use of sold products) includes emissions from your software's execution — every API call, every ML inference, every video stream. For AI-heavy products, Category 11 from GPU inference can dominate the entire emission footprint.
70–95%
Proportion of total emissions that are Scope 3 for typical software companies — making Scope 3 the primary focus of measurement investment, not Scope 1/2 which are relatively small and easier to measure
Cat 1 & 11
The two most material Scope 3 categories for tech companies: Category 1 (cloud compute and SaaS purchases) and Category 11 (customer use of your software). Together these typically represent 60–80% of total Scope 3
CSRD
EU CSRD ESRS E1 requires disclosure of Scope 1, 2, and 3 with audited assurance — for tech companies in scope this means building measurement infrastructure equivalent to financial accounting systems, with audit trails per transaction
Measuring Cloud Compute Emissions (Cat 1)
Each major cloud provider offers carbon footprint tools: AWS Customer Carbon Footprint Tool, Google Cloud Carbon Footprint, Azure Emissions Impact Dashboard. These provide monthly Scope 2 and Scope 3 estimates for your cloud usage. Export data via CSV or API and import into your ESG reporting system. Limitation: provider estimates use regional grid averages and 24/7 matching assumptions — for precise emissions, use Electricity Maps API to get hourly grid intensity for each region and multiply by your actual compute hours.
Business Travel (Cat 6)
Pull flight data from your corporate travel management system (Concur, TripActions/Navan) via API. Apply ICAO or DEFRA emission factors by route distance and aircraft class. Add hotel nights using standard emission factors (0.07–0.15 kgCO2e/room-night). Business travel is typically the second or third largest Scope 3 category for software companies — and one of the easiest to measure accurately from existing expense system data. Automate the calculation with a monthly Python script connecting to your travel API.
Employee Commuting (Cat 7)
Collect via annual employee survey: commute distance, transport mode, days per week in office. Apply transport-mode emission factors (car: 0.17 kgCO2e/km; train: 0.04; cycling: 0). Multiply by commuting days per year. For remote-first companies: home office energy use (electricity and heating attributable to work) falls in Category 7 — estimate as 8 hours/day × home energy intensity × emission factor. This requires employee survey data and becomes material for large, office-based workforces.
AI Inference Emissions (Cat 11)
For AI-intensive products: estimate customer-facing inference emissions. Collect: GPU hours by model size per API call, data centre PUE, grid carbon intensity for hosting region. Calculate: (API calls/month × GPU compute time per call × GPU TDP) × PUE ÷ 1000 = MWh → × grid intensity = kgCO2e. This is increasingly material for LLM-based products — GPT-4 inference costs ~0.0001 kgCO2e per query; at 10M queries/month = 1 tCO2e/month. Connect to your data analytics platform for automated monthly calculation.
GHG Measurement Infrastructure
Our data analytics, DevOps, and software development teams build GHG Protocol-compliant Scope 1/2/3 measurement systems for technology companies. Book a free advisory session.