Carbon-aware edge routing: 6 months of measurements

We deployed an edge function that reads three electrical-grid intensity feeds (RTE for France, ENTSO-E for the EU, NESO for the UK) and dynamically picks the lowest-carbon region to serve a request, provided it's within acceptable latency.

Over 180 days of observation, the gap between the dirtiest region (CO₂ per kWh) and the cleanest regularly reaches a factor of 4 — typically night-time France (nuclear-dominant) versus afternoon Germany (peak coal).

Actual saving on a single visit's footprint is low — 15-20 % on average — but multiplied by millions of requests, it adds up. The experiment's code is on GitHub. Technical takeaway: it costs nothing to integrate if the app tolerates a 50-100 ms latency bump.