| Scenario | vCPU-hours | PUE | Renewables % | Grid CI | Emissions (kg) | Score | Grade |
|---|---|---|---|---|---|---|---|
| Baseline | 12,000 | 1.35 | 65 | 320 | ~3,300 | ~70 | B |
| Greener region + higher renewables | 12,000 | 1.25 | 90 | 180 | ~900 | ~86 | A |
| Low utilization + carbon-heavy grid | 12,000 | 1.75 | 20 | 650 | ~9,900 | ~41 | D |
- Compute_kWh = (vCPU_hours × watts_per_vCPU ÷ 1000) × PUE
- Storage_kWh = storage_GB_month × kWh_per_GB_month
- Transfer_kWh = egress_GB × kWh_per_GB_egress
- Idle_kWh = Compute_kWh × (idle_overhead% ÷ 100) × (1 − utilization% ÷ 100)
- Total_kWh = Compute_kWh + Storage_kWh + Transfer_kWh + Idle_kWh
- Effective_CI = grid_CI × (1 − renewables% ÷ 100)
- Gross_Emissions_g = Total_kWh × Effective_CI
- Net_Emissions_g = Gross_Emissions_g × (1 − offsets% ÷ 100)
- Net_Emissions_kg = Net_Emissions_g ÷ 1000
- Clean = renewables%
- Efficiency = map(PUE: 1.0→100, 1.8→50, 2.5→0)
- Utilization = utilization%
- EmissionsIntensity = map(gCO₂e per vCPU-hour: 0→100, 50→50, 150→0)
- Circularity = hardware_circularity% + bonuses (managed + region)
- DataMovement = map(egress_GB per vCPU-hour: 0→100, 2→50, 5→0)
- Overall = 0.25×Clean + 0.20×Efficiency + 0.15×Utilization + 0.25×EmissionsIntensity + 0.10×Circularity + 0.05×DataMovement
- Enter your monthly vCPU-hours and an average watts-per-vCPU estimate.
- Set your facility efficiency using PUE and your renewable share.
- Add regional grid intensity, utilization, storage, and data transfer.
- Optional: adjust assumptions and offsets if you have better data.
- Press Submit to see the score above, then export CSV or PDF.
What the score represents in operational terms
A cloud sustainability score converts engineering signals into one comparable number. It blends clean energy, facility efficiency, utilization, emissions intensity, circularity, and data movement. The output is built for scenario testing: change PUE, renewables, or utilization and observe the shift. Treat it as a planning metric, not an audit. Tracking monthly helps quantify benefits from right-sizing, autoscaling, and smarter regional placement. Pair the score with service dashboards to avoid shifting impact elsewhere.
Energy drivers you can influence quickly
Compute energy often dominates, so validate vCPU-hours and watts-per-vCPU first. Raising utilization reduces idle overhead and can lift the score without new hardware. PUE reflects cooling and power losses; choosing more efficient regions improves results quickly. Storage and transfer are smaller for many apps, but logs, backups, and media delivery can make them significant. Tune the advanced factors to reflect tiering, replication, and traffic.
Emissions intensity and the renewables lever
Net emissions depend on total kWh and effective carbon intensity. The model reduces grid intensity by your renewable share, so cleaner regions and higher matching create large gains. Compare two regions with similar utilization: the cleaner grid typically scores higher even with similar energy. Offsets reduce net emissions after estimation, but they should not replace efficiency work. Use them for residual impact after demand reduction. When possible, verify claims with contractual evidence and provider sustainability reports.
Using sub-scores to guide engineering work
Use the sub-scores to find the bottleneck. If utilization is low, improve autoscaling, bin-packing, and retire idle instances. If data movement is weak, cut egress with caching, compression, CDNs, and fewer cross-region calls. A low facility efficiency score suggests changing regions, adopting managed services, or consolidating workloads. Circularity improves when you select providers with strong repair, reuse, and recycling practices. Document the chosen levers so teams can replicate successful changes.
Governance and reporting best practices
For consistent reporting, keep assumptions stable and document any changes. Record sources for grid intensity, renewable claims, and monitoring totals so trends stay interpretable. Use tags and cost allocation to estimate vCPU-hours per service. Set targets such as “+5 points per quarter” and connect them to actions: right-sizing, region migration, and architecture refactors. Export CSV for analysis and PDF for stakeholder updates, then review progress in regular check-ins. Add score thresholds to change reviews to catch regressions early.
FAQs
1) Why does utilization affect emissions?
Low utilization means more idle capacity, which still consumes energy. Improving utilization reduces wasted kWh per unit of work, lowering emissions intensity and increasing the score.
2) What value should I use for watts per vCPU?
Use a measured estimate from monitoring where possible. If unavailable, start with a conservative default and refine over time using benchmarking, instance specs, and observed power trends.
3) How should I interpret renewable energy share?
It represents the portion of electricity matched by renewable sourcing or equivalent clean energy programs. Higher values reduce effective carbon intensity in the model and improve the score.
4) Do offsets always improve the score?
Offsets reduce net emissions after calculation, so they can raise emissions-related metrics. However, they should supplement, not replace, efficiency and clean energy improvements.
5) Why include storage and data transfer?
Storage and network activity consume energy in disks, replication, routers, and links. For data-heavy products, these factors can materially affect total kWh and emissions.
6) How often should I recalculate?
Monthly is practical for trend tracking. Recalculate after major changes like region migrations, platform shifts, or autoscaling updates to validate the impact on your sustainability score.