I do agree. There is, however, the problem of fragmentation that is a result of legacy code and the insistence of large corporations to be protective over their IP. As such, fragmentation is as much a consequence of protectionist behaviour as it is of technological challenges.
RISC-V is a large step in the right direction for this run-anywhere aspiration. A large part of the fragmentation is due to hardware incompatibilities. x86-64, and ARM. Based on what I can see, there's a large drive from future-thinkers on RISC-V to overthrow the problem of vendor lock-in and the resultant fragmentation it causes. As such, a project like Libre GPU roadmap could be very successful.
Stated otherwise, there could be 'one ISA to rule them all', and its large-scale success could solve a lot of the fragmentation problem. Based on what I can see from the RISC-V ecosystem. This is exactly what a whole bunch of vendors are planning on doing.