Chip industry contemplates photonic-electronic integration
"So far, the killer application has not appeared, but areas such as medical imaging or sensing may be interesting to watch."–Richard B. Dasher
Photonic discrete devices, such as laser diodes for CD players, have long been part of the semiconductor industry. Recently, advances in optical-network technologies, along with a general concern that electronic devices may reach physical limits of scaling, have led to an upsurge of interest in the use of photonics in integrated circuitry.
Photonic-electronic (P-E) integration can be incorporated in optical interconnects, optical backplanes, the use of optical pulse for clock distribution, optical I/O, and other on-chip and chip-to-system solutions. But how will the integration of this technology from off the silicon microelectronics roadmap affect the dynamics and business models of the chip industry?
It is useful first to review current trends in the chip industry. Four major tendencies can be recognized:
- Most companies are continuing their 20-year shift away from vertically integrated business models toward narrower focus on individual segments of the value chain. This tendency can be seen in the rise of outsourcing to packaging firms and foundries, and in the procurement of standard design tools, design libraries, and other platforms from third-party providers.
- Accordingly, the mainstream chip industry has come to focus on chip-level design (more than manufacturing) as the center of gravity of the value chain. Most companies now bet their product distinctiveness and time-to-market on in-house designs that can be implemented through standard manufacturing processes, especially CMOS. Nevertheless, the industry has not yet developed an active market in design IP (information processing), except for a very few special types such as microprocessors and interface cores.
- Company size in the chip industry is polarizing into larger and smaller units. Large firms are growing larger by actively acquiring other companies for specific technology assets or market strengths. Lower barriers to entry have greatly facilitated the creation and growth of start-up companies. Mid-sized firms, however, face severe challenges of innovating at the same time as they maintain and grow their existing businesses. Mid-sized firms lack the R&D resources of larger firms but at the same time cannot focus exclusively on one new idea, which is an advantage of smaller firms.
- Market applications driving chip industry growth have shifted away from desktop PCs to consumer electronics and other areas such as communications. These applications have encouraged R&D focus on power reduction, analog and mixed signal functions, system-level on-chip integration, and low-cost parameters for any technical solution.
Photonic-electronic integration favors vertical integration over disaggregated business models. These new technology solutions still lack standards for design tools, packages, and system interfaces. Consequently, they will likely first appear in the market in new end-to-end systems solutions, rather than in individual components to be sold in merchant markets.
Photonic-electronic solutions do not support the trend toward value-chain concentration on chip-level design. Manufacturing yield and packaging cost are major challenges for P-E technologies, and so they are likely to be major factors in differentiating product cost as well as performance.
On the other hand, P-E integration does support the trend toward industry polarization into larger firms and numerous start-up companies. Large companies have access to diverse R&D resources and "internal" (affiliated company) customers, an enabling factor in introducing new end-to-end solutions. At the same time, relatively low barriers to entry encourage the commercialization of early stage university and national lab research through start-up companies. Unlike mid-sized companies, such ventures can risk everything on a new solution without having to support prior business lines.
In addition, it is unlikely that P-E solutions will appear first in consumer markets, which prefer reliability and low cost over performance. So far, the killer application has not appeared, but areas such as medical imaging or sensing may be interesting to watch. Such applications require that very large amounts of data be processed at very high speed, and these markets allow higher margins (necessary to recoup R&D cost and fund continued development).
Currently, research into P-E integration is an active topic at major chip companies and also at many universities and publicly funded labs worldwide. Applications of P-E are more likely to appear from this direction than from the optical components industry.
Recent over-investment by U.S. telecommunications firms, however, has slowed (or temporarily stopped) further capital spending in optical network technologies. Optical network equipment and component companies thus currently lack the fiscal resources needed to commercialize P-E solutions.
Some trends to watch include the technology transfer of P-E technologies from corporate R&D (or universities) to business-division laboratories, the development of device and technology development platforms (such as design tools), and the appearance of photonics chip packaging and interface standards. These will reveal not only the direction of target applications but also the speed with which P-E technologies can be implemented, given the real-world constraints of productivity and cost.
RICHARD B. DASHER is the executive director of the Center for Integrated Systems at Stanford University, Stanford, CA 94305; e-mail: firstname.lastname@example.org.