Financial Pressures Meet Chip Development

Decisions about what kinds of chips will be developed are changing, and not for the usual reasons.

While there have always been technological challenges in developing semiconductors—many companies believed that 1 micron was the wall—business decisions are entering into the picture at a level never seen before. Some companies are running their chips for two process nodes instead of one—skipping an entire node in between—and then deciding whether it’s even worth going to the next node or simply taking a half step. The foundries are working furiously to add half-nodes all the way down the Moore’s Law road map.

But for some companies, it’s not even the question of what process node is being used. Complexity has gotten way out of control. Yes, it’s still possible to build almost anything onto a chip. You can add five new functions and myriad power domains that switch on and off like a strobe light to conserve energy. You can add in technologies like air gap at 22nm and use self-assembly techniques, double patterning, EUV (well, maybe) and perhaps even some biochemical processes several nodes in the future.

But the simple truth is that by the time you overcome all of these obstacles, verify the chip and plug all the holes, the money may have been better spent on other parts of the system. Maybe you don’t need to add multiple cores to a chip. Maybe there won’t be software to take advantage of it. Or maybe you’ve found that fixing glitches in software after the chip has already gone to manufacturing works well enough.

A slowing economy and growing competition, primarily based upon price, is only adding pressure to do more with less. It’s not that there will be fewer chips or design starts. Given the role of semiconductors and systems on chips in almost every aspect of life, the design starts will likely increase over the next decade. But the question is what kind of chips will companies be creating? Will they be at the 32nm process node, or will they be at the 80nm node, with a full complement of software?

What do you think the future will hold for chip development? And what kind of chip do you think companies should be developing?