With the semiconductor industry slump heading toward the two-year mark, it is illustrative to scrutinize those companies that are still profitable to see what they are doing right. Two semiconductor industry darlings that have held their heads above water-Intel and Nvidia-have reaped added profits from branding their products to end users.
So, does branding silicon to end users always pay off? Of course not. Branding will always produce extra revenue. However, it is profits, or the branding margin, that companies should care about. If branding costs outweigh extra revenue from branding, a company will have made a costly error.
It makes sense to brand silicon to end users only under a special set of circumstances. Three such sets are silicon architectural complexity, ever-increasing silicon performance unbounded by standards, and low system vendor (i.e., customer) concentration. All are needed to defend a brand against competition long enough to recover branding costs.
Architecturally complex silicon creates a high barrier to entry. The Intel Pentium 4, built in 0.13 micron with 55 million transistors, is the king of complexity. Though similar-performing CPUs can be designed with slightly lower complexity (AMD's Athlon has 37 million transistors), the bottom line is that no more than one or two companies can design and make such a product. Similarly, Nvidia's upcoming NV30 solution (likely a two-chip module) packs a whopping 120 million transistors; only ATI remains a competitor, with its 110-million-transistor R300.
The barrier to competitors' diluting a brand investment gets higher when end users demand higher performance. Though products like Intel CPUs and Nvidia graphics processors have open standard interfaces, they do not have standards governing performance and functionality in the way a Gigabit Ethernet transceiver IC does. Architectural complexity means designing big and complicated chips.
Finally, to defend a brand, there must be a low concentration of system OEMs. Though the chips are complex, the systems that use them are easy to build.
This analysis suggests that there are a lot of chips not worth branding-Gigabit Ethernet transceivers, cell phone chips and DRAM chips, to name a few. Think of some of your own that are worth branding and seize the day!
Jeremey Donovan ( Jeremey.Donovan@Gartner.com) is chief analyst at Gartner Dataquest.
RSS
RSS| 
Digital| 
Mobile
