[Google Blogspot ate my 10 paragraph blog. Honest. Below is a thumbnail of what never got posted.]
HP has by far the greatest risk to Itanium disappointments due to the company's heavy commitment to the processor family. HP 9000, VAX, and Tandem NonStop customers who bought into the Itanium migration are now facing some difficult choices. The delayed Montecito will have reduced features and a lower clock than planned when it arrives mid-2006. Future Itaniums are pushed out nine months or so. Thus, the future for tier-1 backend database machines based on Itanium is cloudy and deteriorating.
HP's Itanium engineers were sucked into Intel last year, meaning HP is stuck with what Intel can deliver on the Itanium platter. There's lots of speculation that Intel, absent its contractual commitments to HP, would sunset Itanium as a distraction. If nothing else, Intel needs the Itanium engineers as troops in its big battle pitting Xeon -- and its architectural descendants -- against AMD's Opteron.
The question I hear HP's Itanium customers asking today is basically this: "When can we jump from an Itanium multi-processor in a scale-up architecture to a scale-out architecture with 4-way Xeon (or Opteron) processors?" The answer is that many can. Today.