Is Intel following an outdated business model?

"No," argues Rob Enderle, the lead PC analyst at Giga Information Group. "The PC market is built on churn. Performance is what historically has driven churn. For a market that's based on churning systems, performance is critically important. If it was to stall, the hardware companies quickly would lose much of their market value."

Enderle agrees that today's applications don't tax the PC processor, so there's no need to replace a Pentium II with a Pentium III to get software to run faster. But he notes that the older systems now being replaced have other bottlenecks, especially slow drives and too little RAM, as well as aging analog components such as drive platters and fans. Even if PC owners aren't motivated to increase processor speed, they need to upgrade three-year-old machines. So they end up getting the latest processor even if they don't need it.

There's no software on the horizon that will make users demand the faster processors that Intel will nonetheless build. (Windows 2000, for example, should run faster than Windows 98 on systems with 128MB of RAM.) Enderle expects such software to emerge, although it's not clear what that software will be. Technologies such as voice recognition, which had been expected to create demand for faster processors, are now bogged down by hard drive and RAM limitations, not by processors. Other technologies, such as on-the-fly background virus scanning and file encryption, have not been adopted in significant numbers, he says.

Enderle's advice to Intel is to keep developing faster processors, but "to spend a higher proportion of the budget looking for software that uses that power." In the meantime, he says, people are "buying adequate"--and finding that they can more than make do with less expensive systems.