So IBM is to stop making servers based on Intel’s 64bit Itanium processor, and concentrate instead on x86 architecture, particularly in combination with its own X3 chipset.

IBM’s move could be related to its growing fondness for AMD’s Opteron chip. Previously IBM has used Opteron as a specialist processor for high-speed applications, supporting a high level of sequential memory access and a low level of random data access, where its HyperTransport technology could be exploited to best advantage. Now, however, IBM is likely to promote Opteron-based System X servers – previously known as xSeries servers – for more general-purpose roles.

Meanwhile, IBM’s decision to abandon Itanium makes sense. It has painful experience of the effects of sentimental attachment to outmoded technologies. In the past, IBM’s love of mainframe systems drove the vendor to dig a multi-billion-pound hole in which it nearly buried itself.

For high-end computing, it seems IBM will now stick with its own Power processors. If IBM thought Itanium had much of a future it would surely have done the opposite – killing off Power and using Itanium instead.

So, in one fell swoop, Itanium is left with only one of the three major vendors in the high-end server market – HP. There are other vendors committed to the Itanium cause, of course, but they sell into niche markets.

Will IBM keep Itanium in its high-performance computing (HPC) systems? If it does, it could, possibly, hold out some long-term hope for Itanium’s return to wider duties when what are now called HPC capabilities drift into mainstream hardware. However, that is a long shot.

What is worse for the future of Itanium is that the likes of IBM and Sun also have large, and growing, services operations. Sun sells a mix of its own UltraSparc- and AMD Opteron-based servers. And IBM Global Services now generates more than half the company’s total revenue. If these service arms do not promote Itanium-based systems when they specify and implement high-end hardware, then Itanium really could face death by a thousand cuts.

If Itanium does not gain more traction in its already shrunken target market – and IBM’s decision suggests that it won’t – this could leave both Intel and HP with a problem. HP is now deeply committed to the processor, but Intel might be reluctant to pay for continued support and development costs. It has already delayed several versions of the chip and recently said updates would be slowed to an annual release cycle.

So, would it make sense for HP, which co-developed Itanium with Intel, to buy out Intel’s interest in the processor and continue development on its own? Such a move might reassure firms using HP’s PA-Risc systems that Itanium has a future.