wbmw, AMD64 does seem to have an advantage in significant high end compute server markets (much to my surprise). The design wins for supercomputers shows that.
However, the meat of the market for AMD is the Linux Sun-replacement box and the Windows servers, I agree. Introductory to mid-level servers. While we can agree that the opportunity for optimizing desktops for Athlon64 will be limited at first to graphics workstations and high-end gaming, the Windows and Linux server markets have far fewer applications which need this optimizing than desktops. Just a few web servers, database servers and network support applications and you meet more than 90% of the need. Much of that stuff will already come packaged with 64-bit Windows or Linux. So a handful of companies (who dominate these markets) is all you need - Microsoft, Red Hat, SuSE, IBM, Oracle, Semantec - the list starts getting sparse after that.
Thus it really won't take much to get general business needs optimized for Opteron, and AMD will have a good head start to get entrenched in this market before Montecito hits the street.
So will AMD64 start looking like 3DNow!, or will Itanium start looking Alpha, a nice idea that never made it? What obsoleted 3DNow! was that Intel came up with their own direct competitor. Without an x86-64 from Intel, Opteron goes unchallenged.
However, Itanium still has the HPQ architectures. It has a guaranteed market, but I am unconvinced that Itanium will ever make inroads as an x86 replacement.
Market Data 
Markets 
AI
