The past few years have seen virtually all aspects of the semiconductor industry undergo unprecedented transformations. Rapid change has always driven—and will continue to drive—the chipmaking world, a place where everybody is hard at work trying to obsolete what they have just created and introduced to the market.

However, current trends represent an unprecedented confluence of many fast-paced fundamental changes that encompass both technology and business and have taken place over a relatively short time. Timing (implementation target dates) and latency (time to objectives) are essential elements in the equations that define and govern these trends, with direct impacts on cost, profit margin, and return on investment. The complexity and multitude of these transformations, together with the need for astronomical levels of investment and innovation, all over a relatively short time, present numerous challenges and create a very-high-risk business environment.

Technology roadmaps have long played a critical role, especially since the 1990s. Yet some unrealistic roadmaps have compounded the
difficulties of timing and latency, disrupting any meaningful planning strategies. Traditionally, technology has been the main engine propelling the semiconductor world but, more than ever before, market forces and economics have emerged as equally critical factors. Of course, technology is still essential for staying in the game; however, a technology-centric focus could become a recipe for disaster.

Therefore, to manage risks and to reduce the chances of failures, a heavy dose of reality must be injected into the decision-making process, and various directions and options must be much more carefully studied. Fact must be separated from fiction, and solid practical solutions distinguished from marketing hype, so that costly and unnecessarily complex, tortuous paths can be avoided. A multitude of recent experiences with difficult technology transitions—such as the failure of low-k integration at the 130-nm node and the premature introduction of 300-mm fab technology—and vexing business challenges call out for a serious reality check.

This feature is the first installment of a new column in MICRO called Reality Check. In this space, I will examine and analyze the major changes, transitions, and transformations that are under way in the semiconductor industry. I will highlight how the intertwined and inextricable nature of technology and business, combined with the realities of the marketplace, must guide the development of aggressive yet practical technology roadmaps along with optimized, business-savvy economic implementation strategies.

Current major technology transitions include the continuing move to copper/low-k interconnect, the introduction of 90- and 65-nm design rules using subwavelength lithography, and the widening implementation of high-volume 300-mm manufacturing. The entirely new copper/low-k interconnect architecture, based on the damascene process, is designed to enhance chip performance and reduce cost. However, this transition has proven to be very difficult and costly, particularly the integration of low-k dielectrics; in fact, low-k integration at the 130-nm node has been generally characterized as a failure. Consequently, the implementation of low-k materials has been pushed out by at least two nodes from the original roadmap plans.

Process integration and yield issues continue to challenge the industry in the 90- and 65-nm technologies. Furthermore, the increased application of advanced process control (APC) methods, especially in critical manufacturing steps, is needed for managing shrinking process windows.

Perhaps the industry's most important and expensive technological challenge remains lithography. Subwavelength lithography has enabled the patterning of feature sizes far below the wavelength of the laser light source. This high-cost endeavor has been made possible through major advancements in tools and materials technologies, such as scanners, tracks, reticles, resists, and metrology. Optical proximity correction, phase-shift masks, and other optical extension techniques have delivered impressive technical results. Yet these achievements have been accomplished at great expense: the exponential rise in mask costs has major implications on the economic viability of advanced low-volume products.

Next-generation lithography (NGL) may address some of these issues, but the immaturity and prohibitive price tag of many current NGL candidates point to even more possible showstoppers in the future. This is particularly true for small and midsized semiconductor companies, whether they're integrated device manufacturers (IDMs) or fabless, that do not command a high market share or have a high product mix.

Another daunting complication is the lack of a real industrywide consensus on which lithographic technologies to adopt. Perhaps the emergence of immersion lithography (IML) will provide an evolutionary, practical solution over the next few nodes. However, this "wet" approach is quite new and unproven in the field: IML was not even being seriously considered up until early-to-mid-2003. The expected success of the immersion alternative may serve as yet another clear example that even successful evolutionary changes require huge amounts of time and effort.

Consistent with historical industry trends, requirements for lower costs and higher productivity have primarily driven the transition to 300-mm wafers. The full factory automation taking place in 300-mm fabs will allow implementation of true facilitywide integrated yield management and APC systems. The 300-mm fabs of the future will be significantly different from those we know today.

Although the move to a larger wafer size makes sense for cost-productivity reasons, the premature transition to 300-mm technology has resulted in yet another expensive and difficult new technology introduction.

Adding to the challenges facing the semiconductor world are a number of new technology transitions on the horizon that will primarily focus on the front end of the fabrication line, including transistor architecture and design. These innovations include the use of new substrates (silicon-on-insulator, strained silicon, silicon germanium, etc.), implementation of high-k gate dielectric/metal gate electrode, nickel silicide, and increasingly aggressive ultrashallow junctions. Most of these technology changes are accompanied by the introduction of new materials, tools, and processes.

Copper electroplating, copper CMP, copper electropolish, MOCVD, atomic layer deposition, flash-lamp anneal, 193-nm immersion scanning lithography, and supercritical fluids are just some of the recent or emerging production technologies. Other new materials and processes will be introduced through novel devices, such as the long-awaited universal memory chips (MRAM, FeRAM, and maybe PRAM). Whatever direction it takes, the industry will again be expected to go through a rapid learning curve with the associated high costs and risks.

There have been many significant changes under way on the business
side as well, including end-market evolution, the rise of outsourcing,
escalating R&D and manufacturing costs, and an increase in alliance
and joint-venture activities. The severity of the prolonged downturn
experienced through the recent cycle has fundamentally affected the
industry, with the potential for more consolidations and the need for new business models. The end-markets for many systems containing microchips—PCs, cellular phones, digital entertainment gadgets, and automotive applications—have been progressively evolving into the commodity consumer domain. Consequently, characteristics of these markets such as average selling-price pressures, shorter product life cycles, consumer confidence fluctuations, and macroeconomic factors will play increasingly important roles in managing semiconductor businesses.

The outsourcing model has led to the gradual disintegration of the
traditional vertically integrated operations and the emergence of the Asia-Pacific region as the world's high-tech manufacturing hub. Wafer processing, test, assembly, packaging, and, more recently, design have all been outsourced, adding strength to the fabless and foundry sectors.

To dampen the effect of the business cycles and to reduce costs and capital expenditures, many IDMs have adopted a "fab-lite" strategy, relying increasingly on outsourcing solutions. China's emergence, both as a budding semiconductor manufacturing center and a voracious consumer of ICs and other chips, promises to have further serious business implications for the global electronics marketplace.

During the course of the year, I will apply the Reality Check to the important issues I've mentioned here, as I try to bring some perspective to the myriad of tough issues facing the semiconductor industry. I hope you enjoy and benefit from these observations and analyses. In the meantime, have a happy and prosperous 2004.

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit

Questions/comments about MICRO Magazine? E-mail us at cheynman@gmail.com.

© 2007 Tom Cheyney
All rights reserved.