SpectraQC

Footprint methodology

An estimate, and how we make it.

What the number is

Every job result shows an estimated environmental footprint — ~X g CO₂e and ~Y mL water. The ~ is deliberate: this is a derived estimate, never a direct measurement. It covers the usage phase of running your job — the electricity the compute drew, and the water and carbon that electricity implies — on SpectraQC’s EU low-carbon infrastructure in Paris.

It does not claim to be a full life-cycle assessment of the hardware. Usage-phase per job is the part that is both defensible and reproducible; broader scope-3 accounting belongs at the infrastructure-operator level.

How it’s derived

From telemetry we already record for billing — the compute profile your job ran on and how long it ran:

energy_kWh = power_model(vCPU, memory) × duration
co2e_g     = energy_kWh × carbon_intensity × PUE × 1000
water_mL   = energy_kWh × WUE × 1000

The power model follows the Etsy Cloud Jewels / Cloud Carbon Footprint coefficients (2.10 Wh per vCPU-hour, 0.392 Wh per GB-hour).

The regional factors — carbon intensity, PUE (power usage effectiveness) and WUE (water usage effectiveness) — are Scaleway’s published figures for the Paris region: carbon intensity ≈ 0.065 kgCO₂e/kWh, a 2024 average PUE of 1.37, and a WUE of 0.014 L/kWh. Scaleway’s figures follow the ADEME Datacenter & Cloud Product Category Rules (a standardised LCA methodology), validated with the consultancy IJO, with energy-mix data from Electricity Maps. We cite this rather than invent our own.

Why the numbers are stable

Each job pins the exact factor set and power-model version it was computed against, at the moment it ran. We never retroactively recompute an old job against newer factors — so the figure on a job from six months ago is the same today as it was then, and a monthly total never silently shifts underneath you.

Values are stored in SI units (kWh, grams, millilitres) and only rounded for display.

For your own reporting

Per-job figures are available on each result and via the API and CSV export, so a post house can roll SpectraQC’s footprint up into its own CSRD/ESG reporting. Treat them as vendor-estimated usage-phase figures with the basis above.

Questions about the basis? Get in touch.