Chips Are Down for EDA, Semiconductor Companies

45nm Pentium dieSemiconductor design tool vendor Cadence Design Systems has given up on buying rival Mentor Graphics when the price became more than it could afford, or borrow. This isn't an isolated failed deal, but an indicator of the problems that the electronic design assistance (EDA) market has been seeing â€" a flat market, entrenched competition, and nothing new on the horizon making chip designers reach for their wallets.

Back in 2005, there were estimations that the EDA market had hit a stagnation point until someone could shake things up. As of this summer, things haven't changed much. Industry revenue actually declined from the first quarter of 2007 to the same period 2008. What kept things from sinking farther into a porcelain basin were start-ups in leading edge technologies. The reason newer categories of tools hold out hope is because they actually address pressing problems for design companies that established vendors aren't adequately addressing.

Even some of the biggest chip manufacturers have had problems moving to new chip technologies, partly because they don't have the tools to do so, and that is potentially throwing a monkey wrench into the operation of Moore's Law, an empirical observation that the number of transistors, a measure of computing power, has roughly doubled every 18 months for years. The results could be a disaster for the industry.

Geometrically increasing semiconductor prowess has allowed two factors that have made the entire industry profitable. Chip vendors could keep adding more features to chip families, always keeping a rationale for maintaining products with premium prices. Electronics companies would often buy the more powerful chips, rather than older models that did less, because they either wanted to distinguish themselves, or they had to keep up with competitors. At the same time, chip manufacturers could make better use of silicon blanks, getting higher yield â€" more chips per piece of silicon â€" and driving their unit costs down.

But designers and manufacturers have been increasingly struggling with newer chip technologies that pack even more transistors per device. As times from design to actual chip manufacturing increase and costs go up, it gets harder to achieve the volume necessary to make the newer chips economically feasible. Also, longer times to market don't work for electronics companies, which tend to handle products whose life spans are only six months, as opposed to the 18 months or longer it takes to get a new chip out.

These pressures are visible when a vendors like AMD consider unloading their chip plants and becoming so-called fables companies in a quest for palatable economics. If better tools don't appear soon, or new areas of development that veer off the drive for ever smaller chip geometries, then consolidation in all areas of semiconductors will become more important as a lot of companies find themselves in trouble.

45nm chip die image courtesy Intel.