With the launch of Itanium 2 last month, Intel began its onslaught on 64bit computing. After the false start of Itanium's first iteration, the firm is now confident that it has the industry backing and momentum it needs to further dominate enterprise computing. But I'm not entirely convinced that Intel has got it right, and I think it is forgetting one vital thing: its roots.

When IBM based its original PC on Intel's 8088 chip two decades ago, executives from other computer vendors queued up to pour scorn on it. What on earth could such a puny toy be used for? The answer was that it brought the benefits of computing to smaller businesses that simply couldn't afford anything else, and had the potential to grow into a useful productivity tool.

But large enterprises are notoriously conservative, and it took them time to adopt the PC. Corporate desktops didn't take off until LAN technology allowed PCs to be connected to the existing IT infrastructure, and Microsoft Windows gave a consistent look and feel to business applications. Thus it was an army of smaller customers that matured the PC until it became acceptable technology for large firms.

With Itanium 2, Intel has put together an impressive high-end chip that ticks many of the right boxes for enterprise requirements. It will handle applications that need huge amounts of memory and it offers a high level of data integrity. According to Intel, the machine-check architecture in Itanium 2 systems will detect and correct even such phenomena as a cosmic ray strike altering a single bit somewhere in the computer's memory.

But for the first time in Intel's history, it has produced a processor that is unlikely to be seen in anything other than a server. Companies such as Hewlett-Packard might be planning Itanium 2 workstations, but these are likely to cost tens of thousands of pounds each. Other than a few very demanding computer-aided design (CAD) users, it is difficult to see who would make such an investment.

By contrast, systems based on AMD's X86-64 technology will cost about the same as current desktops. When 64bit Athlon systems come to market, they will be faster than current machines but will still run existing 32bit applications. And buyers will be confident they are future-proofing themselves, as they will be able to upgrade to 64bit software at a later date.

There is little current need for 64bit desktop machines, but you could equally argue that with current software, there is little need for 2.5GHz systems either. Consumers and small businesses will, I predict, drive demand for 64bit desktops - just as they are driving demand for ever-higher clock speeds. Intel, by concentrating on high-end servers and abruptly breaking with current software, is throwing this market away to AMD.

Of course, this may not matter. Many companies are moving towards thin clients, instead of PCs, on the desktop, so it may make sense for Intel to focus on back-end systems. But many server vendors are also showing a keen interest in AMD's 64bit platform, which looks like being highly scalable thanks to its HyperTransport interconnect technology. And if large firms ever do decide to go for 64bit desktops, they are likely to find this market has been shaped by the demands of small early adopters.

Have your say: contact IT Week

More IT Week Comments