Lowest-carbon feasible region: europe-north1

Verdant evaluated 11 Google Cloud regions for the eu-west user base with a latency budget of 80 ms, then picked the feasible region with the lowest modeled operational SCI. Carbon is a deployment gate, not a dashboard.

Region decision

Decision log

• Evaluated 11 Google Cloud regions for workload 'api' (user base: eu-west).
• Constraints: max_latency=80ms, residency=any, carbon_budget=0.35 g/1k req.
• Selected europe-north1 (Finland, zone FI) at 29.84 gCO2e/kWh, CFE 98%.
• Carbon-unaware baseline (lowest latency): europe-west4 at 194.35 gCO2e/kWh.
• Rejected africa-south1: latency 165ms exceeds 80ms budget.
• Rejected asia-east1: latency 260ms exceeds 80ms budget.
• Rejected europe-west1: feasible but higher modeled SCI (0.20229 vs 0.04774 g/1k req).
• Rejected europe-west10: feasible but higher modeled SCI (0.44130 vs 0.04774 g/1k req).
• Rejected europe-west2: feasible but higher modeled SCI (0.17502 vs 0.04774 g/1k req).
• Rejected europe-west4: feasible but higher modeled SCI (0.31096 vs 0.04774 g/1k req).
• Rejected southamerica-east1: latency 200ms exceeds 80ms budget.
• Rejected us-central1: latency 110ms exceeds 80ms budget.
• Rejected us-east1: latency 95ms exceeds 80ms budget.
• Rejected us-west1: latency 150ms exceeds 80ms budget.
• Carbon gate passed: modeled operational SCI 0.04774 g/1k req <= budget 0.35.

Functional unit: 1,000 successful HTTP requests. Operational SCI follows the ISO/IEC 21031 structure; embodied emissions (0.004 g/1k req, assumption) are reported separately. Data: Google Cloud region carbon data 2025 (annual average, location-based grid intensity) - github.com/GoogleCloudPlatform/region-carbon-info. Latency values are indicative assumptions. This is a modeled, incremental comparison — not a compliance claim.

Endpoints

healthz · region (eu-west) · region (us-central) · region (za, carbon gate blocks) · SCI report

Live API

The signed image from this pipeline is live on Render free tier: https://greenops-status-api-latest.onrender.com (free instances sleep when idle; the first request can take about a minute).

Carbon gate in action · Johannesburg

carbon budget breach: best feasible region africa-south1 has modeled SCI 1.02627 g/1k req, above the budget of 0.35 g/1k req. Verdant does not ship it and does not silently pick another region; it hands a human the trade-offs:

• Deploy to europe-west2 (London): latency budget 80 -> 160 ms, modeled SCI 0.1750 g/1k req (83% lower than africa-south1).
• Deploy to europe-west1 (Belgium): latency budget 80 -> 165 ms, modeled SCI 0.2023 g/1k req (80% lower than africa-south1).
• Deploy to europe-north1 (Finland): latency budget 80 -> 190 ms, modeled SCI 0.0477 g/1k req (95% lower than africa-south1).
• Stay in africa-south1 (Johannesburg) and raise the carbon budget from 0.35 to 1.0263 g/1k req (2.9x) - needs an explicit, recorded approval.

Gate proof · 10/10 correct

Every pipeline feeds deliberately bad changes through the real carbon engine, OPA policies and SLO probe. If a bad change gets through, or a good one is blocked, the pipeline fails.

Verdant's own footprint

Measured from real job durations via the GitLab API; Cloud Carbon Footprint coefficients, runner grid {"us-east1": 579.44, "us-east5": 343.48} gCO2e/kWh (headline uses the dirtier). Payback = requests served in the carbon-selected region until the saving vs the lowest-latency region repays the pipeline's own emissions.

Proof ledger · recent pipelines

Raw evidence (latest pipeline)

evidence-pack.md · gate-proof.json · ci-footprint.json · ledger.json · policy-decision.json · image-smoke.json · slo.json · sbom.cdx.json · pipeline