by John Cowles, Sr. Director of Technology, Analog Devices
The Chips Act’s ambitious and courageous goal to incentivize the onshoring of critical semiconductor manufacturing and packaging technologies is the first step in a long journey to reestablish the U.S. global position in these critical areas. The regular announcements now amount to over $200 billion spread across the nation, according to a recent SIA publication.
This is incredibly exciting and reverses a long trend that dates back to the 1980s, when the official government strategy shifted from direct funding in silicon technologies to letting commercial markets drive those investments. In those days, semiconductor manufacturing was captive within several semiconductor companies, but exponentially increasing costs to reach the next level of performance became unsustainable.
Enter Morris Chang, who founded TSMC in 1987 with its new model of pooling common needs and investing in semiconductor technologies to address a global scale. Opening these technologies to the world created a fertile environment for start-ups and fueled the innovation engine of the last several decades. This bootstrapped operation today delivers around 15 million 12-inch wafer equivalents a year worldwide, enabling electronics from smartphones to cars and radars. Bringing this model back to the U.S. is the obvious way to get the scale needed to provide a secure onshore supply at a competitive cost and drive reinvestments in the next generation of technologies.
Around the same time the strategic decision was made to let commercial forces drive silicon technologies, the government turned its investment towards the more obscure III-V technologies that were maturing in labs and universities and were seen as a keen need for the DoD. The MIMIC program ran from 1987 to 1995 and drove the development of microwave integration, including device and IC processing, semiconductor and circuit modeling and new materials. The program succeeded in advancing the CAD infrastructure, refining epitaxy from magic to science and establishing captive manufacturing sites or foundries at several program performers across the U.S. GaAs-based monolithic microwave/millimeter wave ICs, known as MMICs, started to appear as products in the open market.
Fabless companies like Hittite Microwave developed massive portfolios of MMICs relying on these nascent semi-open foundries. Other III-V technologies like InP and GaN followed GaAs as subsequent programs continued to drive these technologies in the labs. Since these MMICs typically sit at the frontend of many space, military, instrumentation, and communications applications, they set the fundamental limitations of these systems. At the time, the wafer volumes were modest, and costs were secondary. The embryonic ecosystem created by the MIMIC program was able to address the needs of those days.
The success of the MIMIC program in maturing GaAs technologies caught the attention of the commercial world that needed these technologies to provide backhaul for cellular networks, to miniaturize low-noise front ends for home satellite TV units and to increase mobile phone battery life using newer GaAs-based power amplifiers. The scale and lead times required for these applications could not be fulfilled by smaller-scale foundries that were supporting DoD applications. The table was set for a fundamental shift in the supply chain for III-V technologies, particularly GaAs.
Probably not by coincidence, the first successful pure-play foundries were established in Taiwan, including WIN Semiconductor in 1999, among others. Just as with TSMC, these pure-play GaAs foundries used a global scale to deliver world-class technologies, models, lead times, and customer service at a competitive cost. They developed roadmaps for next-generation processes and explored new technologies like GaN and InP. As pure-play foundries, their core business continues to deliver high-quality wafers in volume. In the meantime, the ecosystem in the U.S. remained captive to many of the performers in the MIMIC programs of the 1980s or even shrank through consolidation.
Fabless companies today, and even DoD primes who need capabilities outside their captive fabs, have limited onshore options that provide pure-play foundry services, offer a breadth of technology nodes, deliver fast cycle times, and reinvest in more advanced capabilities. Most of the III-V fabs in America today either impose limitations on what kind of products can be developed, do not offer complete foundry services, or compete directly with their own product design teams.
The lack of open foundries stifles innovation by severely restricting access to start-ups and other fabless design houses. GaAs technology is the natural gateway for even more advanced GaAs, GaN, InP, GaSb, and even diamond technologies. Without the innovation cycle fueled by a healthy GaAs foundry ecosystem, these other material systems will struggle to reach the needed scale. Such foundries are critical to filling the so-called Valley of Death between research labs and production scale.
Considering the billions being poured into onshoring silicon-based technologies, the onshoring of III-V foundry capabilities would come at a small fraction of the cost, but with the disproportionate impact on the overall performance of these critical systems. Part of China’s investments to achieve independence in semiconductors is directed at Sanan IC, a broad III-V foundry with multiple fabs supporting power, optoelectronic, and RF technologies that are critical to telecom, defense, and automotive markets.
We have nothing even close to this scale in the U.S. Let’s not forget that the U.S. led the establishment of manufacturable GaAs technologies more than 30 years ago. It is time to bring them home. It’s not only critical to our security, it’s good business.