Temporal instead of step funtions and Lambda for Infra provisioning

We have orchestration system which will spin up cloud infra VMs from multiple clouds with step functions along with lambda, all the components and tenants details are from RDS. We see some limitations with Steps functions and lambda.

Is temporal would be right solution for this use case? If so do you have any scale value matrix for temporal like how much memory & CPUs needed for billions of workflows?

Can you give more info about your use case please? Believe Temporal would be a good fit for it, but would help to know more to be sure.

Regarding scale, are you looking at self-deploying your Temporal cluster(s) or cloud? What are you looking at as far as load goes (workflow executions / activity executions) per second? What would be the peak values in your opinion?

We would like to try to deploy our own temporal clsuters on any cloud provider K8s. I am not sure about per secod value, basic activity should be provisioning VMs on multi region multi cloud based on schedule.

Just would like to know of you have any metrics value based on available resources like cpu and memory

Approx we can say about 100k+ step functions/lambda in AWS, expecting 10 times more load for future usecases