I agree that it would likely have to be IBM to sponsor such a project.
Hardware, developers, QA, Project management, documentation, etc would be significant.
Could AIs help reduce the cost or cut time lines? Possibly.
I imagine such a team would take several quarters to produce something nearing an Alpha, productizing several quarters more.
Thinking out that timeline, we'd be in to Power11's life cycle, potentially the Solid Silicon chip would be real as Power10 compatible. The market will have further moved into AArch64, with RISC-V maturing towards early viable production systems. PPCs continue in their markets with constant shrinkage due to migrations. I guess it becomes a long term question - Does IBM have plans for Power 12/13/14 bringing the ecosystem through the 2030s? What is the plan to build development resources for the road map?
Perhaps I'm heading into discussion points for another part of the forum.
From my point of view the low hanging fruit is getting Java based applications to work nicely on POWER. My focus has been on Apache CXF, ActiveMQ, Camel, Karaf, Kafka - effectively SOA style services that business use in integration stories. POWER has a place in that ecosystem given its performance characteristics under load.