If I was at HP I'd push hard for a RISC V port and then try to introduce mirco-hpux, specifically designed to fit within the gaps of server racks, sucking minimal power while providing essential OOB and management features.
They could thrive in that space for at least the next decade
Is there anything specific that makes HPUX particularly suitable for use in embedded contexts? You may as well use Linux or FreeBSD if you need a Unix like kernel for embedded stuff, or a specialised RTOS of some kind.
For it's more traditional markets, years ago they did consider porting HPUX to x86 to keep it alive (alongside other options like buying Solaris as it already ran on x86), but abandoned it for the "going down with the ship" option they did take. Stuff about it came out during the HP / Oracle dispute over Itanium support years back: https://www.theregister.com/2012/05/23/hp_project_blackbird_...
They could thrive in that space for at least the next decade