Sustainability reporting

Spotter estimates the electricity, carbon emissions and water associated with your cloud usage across AWS, Azure and GCP.

It works from billed usage (virtual machine hours, memory, storage, data transfer, AI tokens), converts that usage to electricity, and from electricity derives carbon emissions and water use based on where each resource runs.

These figures are estimates, not absolute measurements — cloud providers don't expose actual per-resource power or water use. Coverage is only from usage whose billing unit converts to energy (hours, GB, tokens).

How it works

electricity = usage × per-component coefficient × regional PUE
emissions   = electricity × regional CO2 factor
water       = electricity × (on-site WUE + regional PUE × off-site grid water factor)

Usage is grouped into five categories: compute, memory, storage, networking and AI. Region-specific factors (PUE, CO2, WUE, grid water) are applied per resource; where a region has no specific value, a vendor fleet average is used as a fallback.

Only usage billed in units that convert to kWh (hours, GB, tokens) is included. Services whose consumption can't be derived from such units are left out. Where a service is automatically copied to multiple regions and the count is known, consumption is multiplied accordingly.

Electricity

Representative coefficients (per-region values override these where available):

• CPU is modeled around mid-utilization between a VM's idle and peak draw where known.

• GPU uses the model's rated power from the VM spec, multiplied by the GPU count.

• When VM specs are unknown, defaults are assumed: 1 vCPU + 4 GB RAM (general), 4 vCPU + 16 GB RAM (AI workloads).

PUE (data-center overhead). Applied per region where the provider publishes it; otherwise a vendor fallback is used — AWS ~1.15, Azure ~1.185, GCP ~1.1.

Emissions

Electricity is converted to CO2 using the carbon intensity of the region's electricity grid. GCP publishes per-region factors; for AWS and Azure, Spotter uses public grid averages (sources include the European Environment Agency, EPA and Electricity Maps). When a region or factor is unknown, a vendor fleet-average factor is used (AWS ~0.369, Azure ~0.343, GCP ~0.412 kgCO2/kWh).

Emissions are location-based: they reflect the local grid mix and do not account for providers' renewable-energy purchases or carbon offsets. Those can be applied separately when assessing Scope 3 emissions.

Water

Water combines two parts, per region:

• On-site — cooling water at the data center, via its Water Usage Effectiveness (WUE). Vendor fallbacks: AWS ~0.15, Azure ~0.27, GCP ~1.0 L/kWh, with per-region values where published. Google publishes no per-region WUE table; the only GCP regions set apart are the data centers it states are air-cooled (Montréal, Phoenix, Sydney), treated as ~0 on-site water — every other GCP region uses the ~1.0 fallback.

• Off-site — water used to generate the electricity, via a grid water-intensity factor (EWIF). Per-region where available; otherwise a US grid average of ~3.1 L/kWh.

Water-stressed share. Spotter also shows what portion of water use falls in water-stressed regions — those the WRI Aqueduct 4.0 atlas rates "high" or "extremely high" baseline water stress (≥40% of available supply withdrawn annually). It is a binary regional flag, not a scarcity-weighted volume.

AI usage

AI token usage is estimated from token counts using per-model-class energy coefficients, weighted toward output tokens (input/cached tokens count for ~10% of an output token). Approximate values:

These are rougher than the hardware-based compute figures, since providers don't disclose the model or hardware behind a token.

On-premises comparison

To estimate savings versus running the same workload on-premises, Spotter applies a reference on-prem PUE of 1.55 (Uptime Institute Global Data Center Survey 2024) against the more efficient cloud PUEs above. This is a generic benchmark — a specific on-prem facility may differ.

ESG / CSRD reporting

Spotter supports ESG and CSRD reporting by giving visibility into Scope 3 cloud emissions, energy and water across AWS, Azure and GCP, from the start of the previous calendar year through the current month. The outputs are estimates intended to support reporting and decision-making, not to replace audited emissions accounting.