I have created a new node pool with a custom vm type, n2-custom-16-24576, in Google Kubernetes Engine.

When we had a normal vm type, n2-standard-16, the nodes were autoscale correctly, and almost all nodes was consuming >90% of cpu or mem resources but after changing the vm type the node number will remain the same even if the consumed resources are at 60-70%, so a lot of waste

Any hints if i'm making something wrong?

Thanks