We cut our GCP bill by 35% just by fixing autoscaling — no architecture changes

Cloud Computing & SaaSServerlessCost OptimizationGCP

Our Cloud Run services were scaling to min-instances=1 across 14 services 24/7, most of them barely getting traffic outside business hours. Set min-instances=0 for internal tools and non-critical services and let cold starts happen. Saved about $2,100/month. Lesson: audit your idle floor before you touch anything architectural.