What will convince the market is not just this 1 B investment in inux but rather a formal announcement of EOL for AIX and porting of all the reliability features to Linux. That will provide a fair fight to x86. Actually OpenPower was not really necessary, just commodity availability of Power8 would have been enough. Imagine a 12 core DIY gaming rig with a Power8 with MBs from the usual culprits. ! But this is even better but somebody better come up with low cost Power8 parts. AMD shoudl dump ARM and use Power instead for its alternate server strategy.
RHLinux with KVM is supported on Power8. In fact for RDBMS solutions, Power is the preferred architecture if you do not want Oracle. Licensing costs work out cheaper since per core perf. is higher than x86. Not sure how IVY bridge EP fares. In fact I just advised two major banks on this configuration. And if you go the DB2 route it works out even cheaper.
So Power is far from being dead in various niches in the enterprise market. If we have multiple sources and lower system costs, it has a good shot at x86 since it is better in every respect - faster, better RAS features, better threading and better virtualization support. A better processor than Intel in every respect that did not make it due to IBM's premium positioning.
When OpenSPARC came, the foundry eco-system that exists today did not exist. Much easier to do your own processor now. So the situation is different this time. I am not saying that IBM will succeed becuase of this but that the Sparc fiasco cannot be extrapolated to predict IBM's chance of success. ARM has kind of blown in in scaling up for server apps. Lot of system arch pieces are still missing. Will take a few years for them to get there.
IBM's best chance of sucess, zero cost ISA liecense and no nonsense like ARM's deep arch lic. You pay only if you use IBM's RTL. Nobody would look at ARM in the server market if that happens.
The principals at Ross reorganized as Serverworks, a leading chipset supplier for Xeon-based servers back in the early 2000's. Serverworks' market share soared as Intel tried (unsuccessfully) to push Rambus-based chipsets into the server market, while Serverworks had a SDR/DDR/DDR2 road map. Intel realigned its server roadmap to be more in sync with standard DRAM technology, and Serverworks saw its place in the server chipset market fade away. Eventually Serverworks was acquired by Broadcom, and the key Ross principals moved onto other companies.