Easy actually: because a well-defined restricted subset of HTML can be machine-audited and there is no way to abuse it. Also, Google can save resources at index

Easy actually: because a well-defined restricted subset of HTML can be machine-audited and there is no way to abuse it. Also, Google can save resources at indexing. But AMP isn't a subset of HTML, is it? They've replaced a bunch of HTML tags with their own amp-prefixed variants… IMHO it reeks of vendor lock-in, and could just as well have been made a proper HTML-subset.