Redefining semiconductor industry R&D
10/01/2008
 |
In 1987, Morris Chang, President of Taiwan’s Industrial Technology Research Institute (ITRI) and a 25-year veteran of Texas Instruments, recognized an opportunity to redefine semiconductor manufacturing. Chang’s great breakthrough was to see how the dynamic and highly innovative nature of the semiconductor industry was making conventional IDM-based manufacturing strategies increasingly inefficient.
When Chang proposed the foundry model, it met widespread skepticism???sure, it sounded good on paper, but would a conservative industry with a successful 30-year track record be willing to evolve away from direct control of production? As we now know, the answer was yes. But first, the promise of overwhelming economic advantage had to overcome deep resistance to change and the concern about uncertainty that is so grounded in human nature.
Today, more than two decades into the foundry era, the semiconductor industry is again facing a situation where inefficiencies are pillaging profits. Research and development expenses are growing nearly twice as fast as product revenues, and R&D return on investment (ROI) has become priority #1 for many IC manufacturers. The time has come to assess alternative methods and models that can boost R&D efficiency and effectiveness. With the industry as a whole investing $49B in R&D last year, according to J.P. Morgan and IC Insights, the subject clearly warrants our attention.
For most semiconductor companies, the expected return on R&D investment is to reach market with new and improved products quickly and cost effectively. This is typically achieved through an iterative process of learning cycles in both device design and manufacturing process integration. Significant energy has been invested over the past decade to create EDA systems that make product design more efficient. But there is a new gap in development capability. Until recently, improvements in device performance have come primarily through shrinks. Today, however, innovations other than dimensional scaling, such as novel materials and device architectures, are the primary drivers. And these innovations have become a primary focus of semiconductor R&D.
A simple fact is coming into play when it comes to advancing these innovations from concept to possibility to reality. If you can learn faster than your competitors, you will have a differentiated competitive advantage. This advantage stems from the ability to rapidly gain an understanding of the relationship between process integration parameters for new materials, device structures, and the resulting device performance.
Combinatorial development techniques, which have a strong track record in the pharmaceutical and energy sectors, offer our industry tools to achieve accelerated learning rates. By conducting massively parallel empirical experimentation-which can involve running multiple experimental splits on a single wafer, matching experimental throughput with characterization processes, and automating data analysis-combinatorial processing provides average gains in development efficiency of over 100??. With this elevated experimental horsepower, development teams can perform comprehensive investigations of far broader solution sets. The bottom line is that better high-volume manufacturing solutions are identified faster, at a fraction of the cost of conventional R&D methods.
In the semiconductor industry, as with most industries that are heavily dependent on the core disciplines of materials science and physics, experimental data is the only thing that moves ideas into production. Because we need more data faster and at lower cost, a change is required; a change that will allow us to overcome the economic impediments to sustaining the industry’s historic pace of innovation. If we take the lesson of Morris Chang’s vision, we will get our data, and maintain that pace???and indeed, many chipmaking companies have already begun.
For more information, contact Dave Lazovsky, CEO, Intermolecular, 2865 Zanker Road, San Jose, CA 95134 USA; ph.: 408-416-2300; email [email protected].