Reviving U.S. semiconductor production, at the apex of the electronics industry, is unlikely to succeed without also rebuilding the more basic domestic ecosystem of companies in chip assembly and test, according to industry experts.
While U.S. legislators prepare to approve a $52 billion package of incentives to help revive the domestic semiconductor industry, there are concerns that most of the financial support will go to chipmakers that don’t need assistance while the dwindling U.S. assembly and test segment goes neglected.
“There is discussion right now on incentive programs,” Matt Kelly, the chief technologist at global electronics association IPC, said to EE Times. “There’s a lot of concern about just propping up the industry artificially and when the incentives go away, the whole game plan goes away. Unfortunately, that’s — at least from my perspective — where we are right now.”
One of the biggest gaps in the U.S. electronics industry is IC substrates, which boost chip density on printed circuit boards used in small footprint, low-power devices for automotive, IoT and 5G applications.
“We have never had IC substrate production in North America,” Kelly says. “This isn’t a bring-it-back story. It’s a we-need-to-get-it story.”
The IC substrate market was worth $7.7 billion in 2020 and will grow to $12.2 billion by 2027, according to a report by Mordor Intelligence. Demand for smaller footprint, higher performance and lower power consumption has shifted to materials that connect chips together or connect chips to printed circuit boards in automotive, mobile and IoT devices.
Virtually all the substrate makers are in Asia. The leading suppliers are Unimicron and ASE Group of Taiwan, Ibiden of Japan, and SCC of China. Unimicron has invested more than $700 million in R&D and capacity expansion for advanced flip-chip substrates through 2022, according to the Mordor Intelligence report.
“For IC substrates, we’re over 20 years, maybe 25 years behind Asia,” Kelly said.
Despite strong growth, IC substrates still face a technology challenge from some companies using silicon interposers with multiple redistribution layers (RDLs) for better connections between logic and high bandwidth memory. Others are using fan-out-on-substrate with RDLs. Chip assembly companies such as ASE running flip-chip ball grid array lines need substrates as well as wafer bumping and fab capacity for RDLs. But the fan-out wafer-level packaging group need fewer process steps to the RDL level, and as a result, the industry is shifting toward fan-out technology, according to the report.
It’s on the forefront of technology development where the U.S. needs to invest and maintain electronics industry leadership, according to Carol Handwerker, a Purdue University professor who is developing chip packaging technology.
“What we’re going to need for the future is the new technological advances,” says Handwerker. “It’s not it’s not as simple as it once once was. Right now, it’s heterogeneous integration. It’s hard to tell where the the packaging, assembly and test stops or where the chip stops.”
ASE and Amkor, the world’s largest assembly and test companies, have all their factories in Asia, and they aren’t planning to follow up the recent multi-billion dollar investments of Asian chipmakers like Taiwan Semiconductor Manufacturing Co. (TSMC) and Samsung in the U.S. The scarcity of assembly and test companies in the U.S. will hinder the American plan to build a secure electronics supply chain.
By 2024, when Intel, TSMC and Samsung start to produce more chips in the U.S., those companies will have to put their silicon wafers on a plane and ship them to Asia, he said. “We’re lengthening the supply chain, not shortening it.”
Unlike the more critical issue with IC substrates, kickstarting assembly and test in the U.S. is more about ramping up existing production volume. North America has over 25 OSAT facilities, but they are “mom and pop” operations that are not in the top 20 by annual revenue, Kelly added.
“The ability to get in and to build advanced packaging technologies, pull up our socks and do things in a better way, is a much easier task for OSAT (outsourced semiconductor assembly and test). It’s a lot more money and infrastructure. It’s not necessarily as much about lack of skill and knowhow. OSAT is an easier problem to solve. With IC substrates, we have a giant hole.”
In the meantime, ASE, the world’s largest assembly and test company, is doubling down in Asia. In December 2021, ASE Technology Holding Co. sold manufacturing assets to Beijing Wise Road Asset Management Co., a private equity investor that’s helped build China’s semiconductor industry. Wise Road joined with Beijing Jianguang Asset Management in 2016 to buy NXP Semiconductors’ standard products unit for $2.8 billion, later renamed as Nexperia. The US government in December 2021 blocked Wise Road’s planned $1.4 billion acquisition of South Korea’s Magnachip.
Through its transaction with Wise Road, “ASE will improve its overall competitive edge by optimizing its strategy and resource allocation in China, while further enhancing its investment in advanced technology development and expanding its leading-edge capacities within Taiwan,” ASE said in a statement.
“The purpose of doing this transaction is really to better realign our resources and to focus on the investment in our mega site, which is SPIL Suzhou, to continue to address the China opportunity,” ASE CFO Joseph Tung said in a February 2022 meeting with analysts. “This year, we’re expecting very high growth in our SPIL Suzhou site as a result of revenue coming from the expanded customer base, mostly Chinese customers.”
China’s assembly and test companies are building capacity as the nation creates its own secure supply chain, according to ASE chief operating officer Tien Wu, speaking at the February event.
“I think it’s natural for our China competitors, under the support of the industrial market as well as the government, to build up capacity because the mid-term or even the short-term objective is to create a self-sufficient ecosystem in China.”
OSAT companies, more labor-intensive than chipmakers like TSMC and Intel, are undergoing a technology upgrade.
Last year, ASE built 25 automated “lights out” factories that eliminate assembly line workers. This year, ASE is boosting that number to 37. “We will continue to add 10 lights-out factories per year or maybe accelerate over time based on customer volume,” Wu says.
People in the US also expect assembly and test to become more technology intensive. Peter Bermel, a professor at Purdue University, aims to improve radiation hardness, which will gain in importance as private companies increase space launches. Bermel is working with the US Department of Defense to develop new rad-hard technology.
“We have a big upward trend of the ability to survive space environments,” he says. “That’s going to be integral to everything like SpaceX, Blue Origin and what all these other private sector firms are trying to accomplish.”
When space exploration started more than 70 years ago, almost everything was in low earth orbit. Now, there’s a wider range of potential environments to consider, so understanding those environments and radiation sources will be more important, according to Bermel.
One other task for radiation hardening is the question of survivability in a nuclear conflict. Being able to survive in those environments is a priority for the Department of Defense as well as the Department of Energy and other parts of the US government, Bermel says.
Back on the ground, the US has a lot of gaps to fill. Part of the challenge is educating enough people to expand production.
“We don’t have much of an assembly and packaging business in the United States, except for defense, some aerospace and and implantable medical devices,” Purdue professor Handwerker says. “We have a major Department of Defense workforce development program called SCALE. It’s to provide the DOD with a workforce that has asymmetric training relative to our adversaries.”
The US needs a viable chip ecosystem that’s competitive with Asia, Bermel says. “If you don’t have that, and you just have a small part of the ecosystem, then you have a problem where you basically are dependent on Asia still.”
Lobbying in Washington
Persuading the US government on how incentive money should be apportioned among US government programs as well as private companies in an ecosystem ranging from chipmakers to printed circuit board suppliers hasn’t been easy.
About $2.5 billion of the $52 billion planned for subsidies will probably go to advanced packaging, according to an electronics industry lobbyist in Washington D.C. who spoke on the condition of anonymity.
“The executive branch is still trying to figure it out,” the lobbyist said. “It’s not that the US government has never ramped up a project of this size, but it’s it’s pretty rare, especially in peacetime.”