The logic is that if you outsource your ops team to AWS you have less people but your revenue is still the same, you're just paying AWS instead of salaries.
So if we take the revenue per person number as anything significant then netflix skews that by using AWS.
It would be difficult so no one bothers. How many "people" is monthly use of S3, DynamoDB, RDS, Route 53, etc. equivalent to? Its not easy or worthwhile to estimate. Revenue per employee can be a bit of a vanity metric.
well, the obvious solution is using profit/people rather than revenue/people, isn't it?
EDIT: ah, as another commenter mentioned, it is also true that you might be growing fast with low profit so I guess both measures make sense (or not) in different contexts.
So if we take the revenue per person number as anything significant then netflix skews that by using AWS.
Ultimately it's a number without much meaning.