For music, synthesis and DSP are basically solved problems. Synthesis won't get more interesting if you throw more cycles at it.

Buyers have mostly rejected more complicated kinds of synthesis - including resynthesis and physical modelling - because they don't understand them.

So we're stuck with expanded virtual versions of the old analog synth/wavetable/FM/sampler and studio rack processor models, but with a slightly wider range of oscillator types like granular, and somewhat content-aware processing.

There are areas on the R&D edge, but they're not about the audio.

Changing the subject - one reason Roland stood out was outstanding industrial design. Many of the synths and drum machines produced during the 80s peak are beautiful objects which combined a friendly, fresh, youthful look with a very clear programming model.

https://happymag.tv/the-roland-jupiter-8-the-fattest-synth-i...

http://www.vintagesynth.com/roland/juno60.php

Yamaha tried to do colour coding with the CS series, but didn't quite get it right. So they made the world's most playable and expressive analog poly (CS80) instead.

They took the design hint with the DX7, but never quite hit the same design sweet spot with subsequent synths. And unlike the Roland designs, the DX7 was insanely difficult to program.

https://en.wikipedia.org/wiki/Yamaha_DX7

US synth designers never quite understood colour coding, so the designs were more functional and more expensive, which made them feel more serious, more rock band, less accessible and less fun.