For an instrastructure test plan, I would suggest starting by asking for /use cases/ of the infrastruture. For example, provisioning a virtual server of various OSed with 5 minutes.
From there I would find out about negative cases - what if the private cloud is out of disk, out of memory, etc - what error message comes back? Should service get slower on other VM's that are up and staying up? etc.
Of course, I'm making some assumptions above. Infrastructure is a pretty broad term, and you could be doing something completely different. The point is to figure out how the customer could use the tooling, then create happy-path and 'if things go wrong' path test ideas that we document and discuss with our customer.
Here's the real hint: If you can engage the customer in a conversation on what they want or expect to see, then you're halfway there ...