I have been suggesting recently that Linux could prove to be both boon and bane to Intel. The primary reason Intel needs Linux is Windows. Intel keeps making computers faster, more scalable, and more affordable. As Intel computing power has increased, the price of Intel chips has decreased. That makes Intel-based computers increasingly competitive with higher-priced RISC boxes.
The problem for Intel is that Microsoft keeps making Windows fatter, slower, buggier, and more expensive. That means Wintel-based computing is less competitive with higher-priced RISC architectures than it should be. Enter Linux. Linux offers the speed, stability, and scalability that is lacking in Windows NT. It has a remarkable level of compatibility. (It offers NT file and print services, Novell Directory Services, etc.) And it does all of this at a rock-bottom price.
That's why Intel loves Linux. Linux does a better job than Windows at elevating Intel computing into a more competitive position with higher-priced RISC workstations. The problem for Intel is that Linux also runs on these alternative RISC systems and more. It runs on Sparc, the Alpha, StrongARM, Macintosh - heck, it even runs on a PalmPilot. Any vote for Linux promotes an OS that is essentially platform agnostic.
That doesn't automatically constitute a threat for Intel, though. Whether or not Linux eventually spells trouble for Intel depends on a phenomenon I will call "the threshold of fear". The threshold of fear is the dividing line between two kinds of IT customers - those who find the best solution for a business problem, and those who redefine the business problem to fit their most familiar frame of reference.
The most familiar frame of reference for IT departments that grew up on PC hardware is Intel. Anything outside Intel is located (along with other mysterious systems, such as Unix and MVS) on the uncharted part of the map that says "Here be dragons".
Those IT departments that grew out of a heterogeneous mixture of hardware and software platforms have a higher threshold of fear. Their map is much bigger, and they are less intimidated by uncharted territory. If the best solution is an AS/400 and they don't have anyone on board with AS/400 expertise, they hire a cartographer.
Threshold of fear
By the way, it is this very same threshold of fear that explains why Windows NT never amounted to much on MIPS, Alpha, and the PowerPC. Customers who were familiar with alternative architectures to Intel had developed a high enough threshold of fear that they were not intimidated by the operating systems native to these platforms. To put it bluntly, most people who knew their way around established RISC systems knew enough not to run NT on them.
The threshold of fear leads me to speculate on at least two possible futures for Linux and Intel. Linux may simply invade the space below the threshold of fear. This is likely to happen if we see continued Windows 2000 delays, or if it ships and doesn't meet expectations. In that case, Linux may replace Windows as the safe choice on Intel.
On the other hand, there is the remote possibility that continued success for Linux will raise the threshold of fear. In this future, Linux is successful, but fewer people are intimidated by non-Lintel platforms than they were by non-Wintel platforms. If this is the way things turn out, Intel will need to come up with some compelling reasons why its architecture is preferable to the cleaner, more scalable RISC architectures.