It sounds like the technology hasn't evolved much in the last 20-30 years then.
Around 1995 I was involved with a project where an engineer designed and built a multi-kW CW (I don't recall the exact value) pre-amp out of transistor based modules. There was a plan to apply the same approach to a later project at much higher power levels (100's of kW's) where the conventional approach would have been to use a klystron. I left that team (and the field) though so I don't know if the larger version was ever successfully implemented.
EDIT: I should note that this wasn't in the microwave regime, it was 500 MHz.