I have always hoped and assumed that it is not - that set theory (or something) provides a shortcut to the result.
I have created a non-updateable view that aggregates data from several tables, in a way that produces an exponential number of records. From this view, I query one record at a time. Because the underlying dataset is small, this technique works well - but I'm concerned it won't scale.
I've heard MySQL uses temporary tables to implement views. My heart lurches at the thought of potentially massive temp tables popping into and out of existence for each and every query.