wbmw, thanks. (chipguy, your points addressed here also.) On wbmw's specific points -
I really doubt that an executive from IBM would be misled by Intel's positioning of IPF.
I agree. However, top executives at most companies tailor their message to their audience. If the server audience believes that Itanium repeatedly gets repositioned to a smaller (higher) segment then the IBM executive is not going to insult his listeners by raining on their beliefs.
As for the market shrinking, I responded to this in an earlier post on this forum. I think the RISC market is due for a comeback...
By 'market shrinking' I was referring to the repositioning of Itanium by Intel. In the 90s it was to be the x86 replacement everywhere. As Xeon and Merced were released that was changed to server x86 replacement. At the IA32-e unveiling, it was changed (again by Intel) to be 'high end servers', while IA32-e would be used for commodity servers. All along Intel says that Itanium will eventually be the x86-replacement, but the practical (near-term) positioning keeps on changing to a smaller market.
Referring to 64-bit migration of x86:
This is largely irrelevant. Intel has reversed their position on x86-64, but they have not eased off of their IPF positioning in the least. The market seems to be suited to both architectures.
It is my position that the market will be suited to both architectures for only a few years. I project current trends to predict that x86-64 will be a better (smaller, cheaper, just as fast, just as complete) solution for any RISC (or EPIC) system design by the end of the decade.
The IBM model of having a single architecture from top to bottom (Power4 to PowerPC) will be implemented at Sun; it is rumored that Sun is working with AMD to improve cache coherency for future product and I expect Opteron successors to progressively adopt more advanced features to be useful for big iron. Thus IPF and AMD64 are moving to functional equivalence over time (I do not buy the proposition that IPF will distance itself from x86 and RISC). IPF's problem is that few people believe it will be able to compete with x86 in the commodity realm, but x86 will compete with IPF in big-iron in a few years.
A related point that has been discussed here is whether the main server (and software) companies can afford to give up their big-iron revenues. It is the trend (since the early 70s) to put big-iron on a diet. No longer are we talking about $10M IBM-370 systems. Over time all the major server manufacturers have been adjusting to the new world where the price points (and features) of PC servers vs. big-iron get closer and closer.
Software vendors who follow a big-iron model cannot simply implement commodity versions of their stuff, it is a whole different sale. The $50K software package is a small volume, high support item whereas the $500 software package is high volume (and 'support' is often just a mispelling of 'insult'). The corporation just cannot use commodity $500 software for its critical operations, but the $50K software model is being squeezed. I expect that the big-iron model at the end of the decade will be that the software is distributed electronically at virtually no cost and the corporation buys annual service contracts and/or per-seat licenses. Thus the dilemna of purchasing a $50K application for a $5K server will be resolved.
IBM is well on its way to this business model.
Market Data 
Markets 
AI
