The numbers I've seen suggests Intel ships around 300-400 Million processors per year[1]. In 2014, 12 Billon (with a B) devices with ARM cores shipped[2]. In 2011, around 500 Million MIPS based cores shipped[3].
These are obviously very different businesses. But it's clear there is an enormous market for licensable low-end IP cores that are used in everything from cars, cable boxes, TVs, cellphones (many phones have multiple cores internally to operate functions like the cellular baseband), etc. Most of the licensees are just gluing together licensed cores and contracting foundries like TSMC or Global Foundries to build them. The margins tend to be pretty thin (we're talking chips that are in the single digits for cost), so saving a little money on IP licensing is attractive. Since these are often custom applications, binary compatibility is less of an issue.
That seems like a more interesting market for an open ISA and open core processors.
[1] http://www.ft.com/cms/s/0/beff6e56-53ed-11e4-80db-00144feab7... [2] https://atlas.qz.com/charts/Ek18VmbP [3] http://www.forbes.com/forbes/2011/0509/technology-mips-sande...
Maybe these RISC-V chips will never get out of the labs of Google, but they could still serve to force Intel to keep prices the same on its new generations in the future, unless they want Google to really get serious about making its own chips.
Have customizable cores to add HW accelerators to like Cavium's Octeon III's do.
Things might get cheaper over time as big vendors buy in volume and that money goes back into enhancements.
I don’t see how they’re going to match raw performance against those 1tflops+ 20-cores xeons. I don’t see how they’re going to match performance/watt against 7w quad-core C2338. I don’t see how they’re going to match price against $20 x5-Z8300. If none of the above, then why should Google/HP/Oracle/anyone else buy that?