It is now clear that we are going through the worst semiconductor industry recession ever. Yet, we are told, the industry is bound to resume its growth curve as it did many times in the past. I am afraid it might not.
Let's examine whether the driving forces behind industry growth still exist; if they don't, that growth won't resume. For 30 years semiconductor growth has ben driven by the exponential improvements in the cost/performance ratio of semiconductor devices, a phenomenon known to us all as Moore's Law.
New levels of semiconductor cost/performance enabled new products, whose success in the market has been the driver for the industry's phenomenal growth. Calculators, cell phones, digital watches and PCs are cases in point. For each such new product, the classic "S" curve took place: After an initially slow start, improvements in cost/performance enabled higher penetration and rapid market growth, and eventually the ever-increasing cost/ performance drove down the total market value associated with these products.
In its infancy, the semiconductor industry served the aerospace and military markets, which could afford any price. As cost/performance improved, the industry was able to penetrate the industrial markets and create the computer industry. Each level of improved cost/performance opened new and larger markets. But we might be facing the end of this trend.
Currently, the semiconductor industry is driven by the largest market on earth-the consumer market. There is still room for growth, by adding world markets that have been less active in the past and, possibly much more important, by adding new products that will become tomorrow's killer applications.
So where is the stumbling block? Can't we just use the next semiconductor process to develop new products?The answer is no. Not if the development cost of each new, innovative product is over $10 million. Dataquest reported that the number of new ASIC designs dropped from 11 ,000 in 1997 to about 5,000 in 2001. Other analysts reported even lower numbers for 2001.
Usually, it's not easy to prove that large markets await new product ideas. We know the story of the market assessment for Xerox being about 12 copy machines worldwide, and similar prediction patterns repeat for almost every new product. Furthermore, sometimes it takes a few generations to turn a new product into a "killer app."
It's easy to guess which design types were scrapped last year, and which were not. You won't find many "wild ducks" among the 5,000 designs that were done last year. It is alarming to learn that last quarter the chip industry shipped almost as many devices (90 percent) as at the peak time of 2000, and yet we are in the deepest recession our industry has seen.
If the cost/performance improvement of semiconductor processes at 0.13 micron and below is to drive the next level of industry growth, many new products need to be developed. We need to find ways to propel design activity to the historic 10,000 designs a year and beyond. Apparently, the mask-set cost and the associated CAD tool and design costs are only going to increase, and therefore a paradigm shift is called for. The answer could be new ASIC architectures, like platforms, incorporating configurable-logic arrays that use many generic masks, such as ISSP, RapidChip and eASIC, and maskless fabrication based on direct-write e-beam.
For a paradigm shift to happen in a rather short time, a wide industry alignment is critical. This is not trivial, but I believe it's possible. I call upon all the players in the semiconductor industry-equipment makers, foundries, CAE providers, semiconductor vendors and the large system companies-to align themselves with innovation-enabling technologies.
Zvi Or-Bach is the founder, president, CEO and chairman of Easic Corp.