No such plans at the moment. Last time I checked, Gradle didn't support parallelism, and that's a non-starter for AOSP.
Any change in build system is a very massive undertaking. As measured within Google's source tree, the build system is close to 600k lines of code. On its own, re-writing 600k lines is not a trivial task.
On top of that, we have to take the whole ecosystem into account. Even if Google somehow managed to do that switch internally, the rest of the ecosystem would probably have a hard time switching at the same time, so we'd still need to be able to manage classic product configurations and module definitions, for a transition period that couldn't last for less than 2 full releases but might realistically need to stretch over 2 or 3 years.
Since I have first-hand experience in doing massive build system switches (I was involved in one a few jobs ago), here's what my advice would look like:
-tighten the syntax for product makefiles, board makefiles and module makefiles by having the build system issue warnings for the ones that don't follow the rules.
-tighten more by making non-compliance an error but allowing each makefile to opt out.
-tighten more by making the opt-out part centrally controlled.
-move to nested build systems, where the classic build system invokes the new one for all the properly-written makefiles.
-move the special cases over, one by one.
-retire the old build system.
The cornerstone is to tighten the product definitions and board definitions, since those will need to be read by both sides.
The real pain isn't about all the regular makefiles: converting those is just a long slow grind. The real pain is about all the special cases at the heart of the system: libc.so doesn't follow regular rules, neither does core.jar, neither does framework.jar, neither does libwebcore.so, just to name a few "interesting" cases.
JBQ