This is 26 euros because I set a min instances = 1, so there is an idle instance for low latency. If you are ok with 1 second cold startups then you can let it scale to zero for 0 cost when not in use: https://cloud.google.com/products/calculator/#id=95f41bc0-8154-4faa-82ca-a2a5097f1788

Hopefully not all 5 microservices are latency sensitive, if so, why split them into microservices and paying network overhead. Bundle them into a fat service