NSA makes a distinction between data and metadata, as we know. If you assume that distinction, then the question is -- "does the NSA collect actual communications content on millions or hundreds of millions of Americans" -- something which, to our knowledge, is not the case.
Can it? Potentially yes. But we haven't seen anything that suggests that it actually does.
At best, his response contains an assumption which is arguably normal for an NSA director to make, and which may have be conveniently advantageous to hold. But I'm sure, even if you ask him today, he will say that metadata does not constitute intelligence collection.
To prove perjury you need to demonstrate mens rea (guilty intent) and I think it's plausible that he did not intend to mislead the investigation. It was a hard question to answer, because he could not answer in a way that would reveal the existence of the program to collect metadata either.