It may seem wierd, but it is entirely rational.
Sure the ISA is open, but that is just for the CPU. A meaningful inplementation needs all the stuff that goes around it, and, as with all electronics, volume is king.
Theoretically, as you say, someone who needs a CPU to embed could choose Sparc. Then they could set about developing the rest of the system. But when they place an order, they better have a vlome market - or they would be better of with an alternative by a very large margin.
The existing Sparc targets a very specific market (web/database servers) at which it excels, but the market is not really big enough for other players to have massively bigger volume. The machines for this market have more IP outside the CPU than in it - it is about transactions per second, not instructions per second.
I have tried using Sparc as a workstation, and I am using Intel now. Its about the external infrastructure, not the product. My servers are all Sparc (OpenBSD, not Solaris - no hideous licencing problems, and Solaris majors on features I don't need - but if I did, the licence fees might be worth the money.
Now if a Sparc product was to target the mobile phone market?