Information Age: News, analysis & insight for IT & business leaders

When Intel announced that it was introducing 64-bit extensions to its 32-bit Xeon server chip, it not only represented a tacit admission of defeat by the otherwise invincible microprocessor giant, but also raised a number of questions about the future of its flagship 64-bit Itanium microprocessor.

For almost a decade, Intel has been touting Itanium as the chip for 64-bit computing, the one that will blow away proprietary 64-bit architectures such as Sun Microsystems' Sparc and IBM's Power in the market for high-end servers.

Towards the end of the decade, it would become cheap enough to percolate downwards into home and business PCs, enabling Intel to dominate the market from top to bottom with a single microprocessor architecture.

That, at least, was the grand plan, but it has not quite worked out that way.

First, the Itanium limped into service at least three years late and its first version performed particularly poorly, putting off many potential early adopters. Even the Itanium 2, delivered a year later, offered only modest performance improvements.

Second, because of the complicated nature of the design, engineers have struggled to build software that can take full advantage of Itanium's features. Indeed, software developers have complained about the difficulty of writing applications for the chip.

But most embarrassing of all was the success that Intel's main rival, Advanced Micro Devices (AMD), achieved when it delivered its 64-bit alternative last year.

Instead of building its new chip on trendy, but complex, technology, as Intel had tried to do, AMD engineers reasoned that it would make more sense to build a 64-bit chip that could offer backwards compatibility with existing microprocessors.

That meant that as users shifted from 32-bit to 64-bit AMD machines, they would not have to throw away their old applications and start again. They could simply run the old software on the new machines until they were ready to migrate.

The simplicity of the AMD design also makes it cheaper to manufacturer - it is even appearing in PCs costing as little as GBP699.

Intel's response has been simply to copy the strategy of its smaller rival. CEO Craig Barrett even promised that Intel's chip would be "AMD compatible" - a complete turnaround from twenty years of chip-making, when it was the other manufacturers that would claim to be "Intel compatible".

But where does that leave Itanium? According to analysts, it is likely to remain little more than a niche architecture for some time.

For software developers struggling to build applications on the platform, that uncertainty could be fatal. Why invest so much time and energy on a platform that risks being overshadowed and sidelined by a more enticing alternative?