In the mid 1980s, reports of exciting progress from GaAs integrated circuit (IC) performance from R&D laboratories world-wide portrayed a rosy future for GaAs. Now, in the early 1990s, true to their reputation, GaAs ICs are still largely the stuff of the future. In fact, deployment of GaAs ICs in real systems has been disappointingly slow. In 1985, the commercial GaAs IC market was forecast to reach $800 million by 1990. The actual figure was only $142 million. To put this number in perspective, it represents less than 0.4% of the total Si IC merchant market.
In a recent survey of the GaAs industry, Kato explored the causes for GaAs troubles, with startling findings. The issue certainly does not seem to be a performance one because GaAs ICs are sufficiently ahead of alternative technologies. Material quality is not a problem either. Extremely high-quality 3 in. and 4 in. GaAs wafers are now on the market at reasonable prices. On the other hand, several serious deficiencies center around IC manufacturing. The price of the final GaAs ICs is perceived as not competitive with alternative technologies. This is rooted in the low yields and poor repeatability of the manufacturing lines. A great contribution to cost is time-consuming functionality testing, particularly for analog products. For MMICs (monolithic microwave integrated circuits) in particular, final testing can easily become the bottleneck of the entire fabrication process. There is also much uncertainty about reliability. This might explain to a large extent the low customer confidence in the technology. Kato reports that fundamental technical problems in making GaAs ICs are still believed to remain.