ASIC development teams also implement programming interfaces, but there's no guarantee that these match any common industry standards. It's less likely that open building blocks of low-level software will work with ASICs; there typically will need to be at least some software modification to work with custom chips. All of this tends to lengthen development cycles and adds expenses that will be passed on to customers.

Merchant silicon developments usually begin early, as standards are maturing (standards bodies try to finalize details that impact silicon development months ahead of the full standard, to ensure merchant chips are ready). Product vendors can begin designing in merchant silicon months before the chips are ready, based on advance information from the developers. While ASIC developers could take the same approach, it's more common for ASIC development to start later and for products using ASICs to be built serially, further lengthening the time to market.

Since ASIC developments are expensive, companies that build custom chips will work to get the maximum life from them. This can lead to slow refresh times: If companies can eke additional years out of their initial ASIC investment, it makes better business sense to do so even when the platform is falling behind the market.

As the old adage goes, there's a time and a place for everything. ASICs may be a good choice for new and specialized markets, but in the large and mature enterprise network infrastructure market, merchant silicon is the clear winner.

Wilde is senior director of Global Product Line Management at 3Com.