The semiconductor industry is in its fifth decade. Some people say it's a "maturing industry," with fewer challenges and opportunities ahead. They're dead wrong. Our industry has been driven in a virtuous cycle of unprecedented price/performance improvements, stimulating new applications, followed by revenue growth and more R&D investment to further drive price/performance. Still, we have to ask ourselves, is this cycle being broken at its core? Have we reached the end of technology advancement — the heart of the price/performance improvements we've seen?

Clearly the technology advancement game is changing. Simple scaling (shrinking) of MOS devices can no longer produce performance improvements. The simple reason is that we have left the "micro" scale and have entered the "nano" scale. Transistor channel lengths are tens of nanometers wide and gate dielectrics are a little over 10 nm thick. At these atomic dimensions, the transistor "switch" can no longer be turned off completely and high leakage currents ensue. So as we drive more density, chips grow dramatically in power consumption (heat). Have we reached the end? The simple answer is no. We just need to change our approach to the problem.

New materials and device structures will have significant impacts. Many options have already been implemented, such as copper interconnect, silicon-on-insulator and, most recently, low-k interconnect dielectric. Device performance is being improved through sharper junctions, lower capacitances and novel techniques such as strain, to increase mobility within the device. But perhaps one of the most promising opportunities is high-k gate dielectric material, which is being pursued widely throughout the industry.

But these are all still somewhat logical extensions to our basic devices. We need to think in entirely different dimensions. For instance, we need to better optimize our designs. We can pull large memories on-chip through embedded DRAM technology. We can implement "voltage islands" to power down sections of the chip not in use. We can integrate further, through multiple processor cores per chip.

Finally, we need to assume that eventually all of these techniques will run out of steam. We must invest now to create the next nanoelectronics innovations. We need to gather the resources across industry, academia and government to re-create our industry in time for this transition a decade or so from now. Will we meet these challenges? I believe we will and that the ensuing improvements in price/performance will unleash a whole set of applications that we cannot even dream of today. After all, who would have envisioned IBM's 360 computer and how it revolutionized industries from banking to insurance when the first transistor was built? Who would have envisioned the PC and the personal productivity it drove, at the advent of the microprocessor? And who, just a decade ago, would have envisioned the ubiquitous cell phone/PDA and how it has untethered us all and allowed us to always be connected?

And it doesn't end there. We're already seeing signs of how supercomputers are revolutionizing medical and drug research, and how electronics are transforming the entertainment industry, including digitally produced movies and videogames. In fact, it's hard to think of an industry or aspect of our lives that semiconductors aren't transforming as we speak. If there's one thing that we've learned over the past 40-plus years, it's to never underestimate the innovation that our colleagues are capable of generating to drive our industry — and our customers' industries — forward.

John E. Kelly III, Senior Vice President and Group Executive, Technology, IBM Systems & Technology Group, Yorktown Heights, N.Y.