Kinda wonder if the best approach for them at this point would be to target RISC-V and worry only really about running in a hypervisor, or very narrowly targete

Kinda wonder if the best approach for them at this point would be to target RISC-V and worry only really about running in a hypervisor, or very narrowly targeted open hardware, instead of futzing with implementing a pile of device drivers. Make it useful in that environment, and then go from there. Hell, wouldn't it be cool if there was an official GNU GPL'd RISC-V board, running GNU Hurd?