Over the past two years, microchip shortages highlighted America’s reliance on foreign suppliers, and spurred passage of the CHIPS Act, which earmarks tens of billions of dollars as incentives for chipmakers to build new fabrication plants in the US.
But remaking well-established manufacturing supply chains — notably in Asia — is a Herculean task, and “companies are loathe to do it,” according to Yossi Sheffi, an MIT professor of engineering and director of the school’s center for transportation and logistics.
Sheffi believes the US won't likely reduce its dependence on other countries for chip manufacturing — particularly China and Taiwan. “The [CHIPS Act's] objectives are laudable, but from a supply chain perspective, the chances of it persuading companies to move their manufacturing operations back home are very low,” he said.
For one, semiconductor makers have big investments in offshore locations, and reshoring their manufacturing to the US could take decades, said Sheffi, author of the new book The Magic Conveyor Belt: Supply Chains, A.I., and the Future of Work. In addition, moving production to the US could also shift the risk of disruptions.
“Shifting everything to the US will increase some risks, as all the eggs will be in a single basket,” he said. “It will lower the risks of dependency on other countries, especially hostile ones. Thus, a combination of reshoring and friends-shoring may be best.”
The US, where semiconductors were invented, produced 37% of the world's supply of chips as recently as the 1990s. Only about 12% of all computer chips are produced domestically now.
In 2021, the decline in domestic chip production was exposed by a worldwide supply chain crisis that led to calls for reshoring manufacturing to the US. With the US government spurring them on, the likes of Intel, Samsung, Micron and TSMC unveiled plans for a number of new US plants. (Qualcomm, in partnership with GlobalFoundries, also said it would invest $4.2 billion to double chip production in its Malta, NY facility.)
Last August, President Joseph R. Biden Jr. signed into law the CHIPS and Science Act; it provides the US Department of Commerce with $52.7 billion for a suite of programs to “revitalise” the US position in semiconductor research, development, and manufacturing. The first round of incentives totaling $39 billion became available in February.
Essentially, the CHIPS Act is an attempt to increase the percentage of microprocessors produced in the US by closing the cost differential with countries such as Taiwan, South Korea, and China. Thoe nation's governments are already subsidising semiconductor manufacturers.
The five major semiconductor producing nations — China, South Korea, Japan, Taiwan, and the US— are all also top consumers of the technology. (China mainly manufactures mature chip technology, not the most advanced microprocessors.)
Microchips are found in everything from industrial equipment and data centers to cars, smartphones, and gaming systems. Over the next decade, the biggest industries for semiconductor revenue will be servers, datacenters and data storage. They're followed by smartphones industrial electronics, the automotive industry, and personal computing.
“The most challenging factor is the need to move the entire supply chain, not just the final manufacturing stages, but [also] sources of raw material, expert suppliers, leading equipment makers, etc.,” Sheffi said. “The manufacturing of any product, at scale and at competitive quality and prices, requires entire ecosystems, or clusters, of multiple companies and it is a long-term Herculean effort to move everything.”
Another hurdle: skilled labor shortages
Other challenges include the significant upfront costs of building new facilities and creating domestic supply chains to compete with offshore facilities. Additionally, there is a global shortage of skilled labor to run the fabs, including in the US.
One way of addressing that skills shortage would be to revamp immigration laws to make it easier for skilled labor from other nations to work in the US, according to Sheffi. “In-house solutions have not worked well in the US, in part, because of companies’ reluctance to invest significantly in a mobile workforce,” he said.
One important provision in the CHIPS Act that is promising is $13.2 billion in RD and workforce development, and $500 million to provide for international information communications technology security and semiconductor supply chain activities, said, according to Bob Johnson, research vice president at Gartner. "...There’s lots of money for training and developing manpower,” he said.
Microprocessor makers have been working to encourage students to enter the field. “Intel’s been doing that with community colleges in the Portland area and in Arizona for two years now," Johnson said.
And Micron and the state of New York said they will spend $500 million in community and workforce development with a focus on disadvantaged populations over the duration of the plant construction.
The other big, but often overlooked provision of the CHIPS Act, is $1 billion set aside for research into small-scale nuclear power plants to provide electricity to US fab foundries.
While the CHIPS Act sets aside billions in incentives, estimates suggest that funding a new generation of chip manufacturing in the US could cost $110 billion over the next decade, according to the Brookings Institute. Two new plants being built by Intel in Arizona cost about $30 billion alone.
A typical leading-edge microchip fabrication plant costs about $15 billion dollars to build. That's more than the $13.3 billion cost of the Gerald R. Ford aircraft carrier, the most advanced ship of its kind currently under construction. It's expected to take three to four years to complete. In comparison, manufacturers such as Intel and Samsung can start with a bare field and have a fab plant up and running in about two years, Johnson said.
Is the CHIPS Act having an effect?
So, did the CHIPS Act play a role in semiconductor manufacturers committing to new facilities in the US? The answer is likely no, Johnson said.
The announcement of new semiconductor fabs by Intel, TSMC, Samsung, Texas Instruments, and others all came before the law was passed. The only exception is Micron, which in October announced plans to spend $20 billion to build a fabrication plant in New York, Johnson noted. Micron said it may also spend up to $100 billion over the next 20 years on expansions to the facility.
In September, Micron also broke ground on a memory manufacturing facility near its headquarters in Boise, ID. That plant was Micron’s first new such facility in the US in 20 years.
Government incentives and investments in American semiconductor chip manufacturing, such as thosein the CHIPS Act, have been a part of the semiconductor industry for years, according to Johnson. “You’ve got a lot of politicians out there talking about wanting to bring advanced semiconductor manufacturing back to the US. And I think that’s great, but it’s been here already for quite some time,” he said.
What’s likely incentivising semiconductor makers more than government dollars are geopolitical changes. Taiwan is potentially a major choke point in any electronics supply chain. Any electronic part, whether for a smart phone, a television, a home computer, or a data center likely includes critical components that came through Taiwan.
“If you look across the Taiwan Strait, you’ve got this 900-pound gorilla called China that is saying 'Taiwan belongs to us, and if you won’t give it to us, we’ll take it at some point,'” Johnson said. “What would happen to the semiconductor industry if TSMCs fabs were destroyed? Disaster.”
Before Chinese President Xi Jinping became president in 2012, Western nations had a relatively healthy trade relationship with China. Since that time, it has become more contentious.
“Before Xi came in power, we had this great trade relationship. And there was the belief that if you treated China like a grown-up partner, they’d start acting like one; that turned out to be a very bad assumption,” Johnson said. “So yeah, the idea of bringing the entire supply chain back to the US? Probably not practical.
"But you want to figure out how to diversify away from China as much as you can. I don’t consider China a reliable business partner anymore.”
The most advanced semiconductors come from three companies, Intel, Samsung and TSMC, which stands for Taiwan Semiconductor Manufacturing Company. But advanced microchip technology, including the equipment and know-how for making new microprocessors, has been dominated by the US, Europe and Japan over the last 30 years, Johnson said. In the US, Intel, Nvidia, and Qualcomm dominate advanced chip layout and architecture.
While TSMC may be the world’s largest contract manufacturer of the semiconductor chips, that’s because it’s a foundry for other companies’ technology. In other words, companies such as Intel, Qualcomm, Nvidia, and AMD design and test the latest microprocessors — then contract with TSMC to manufacture them.
“Intel promotes chips. TSMC promotes manufacturing, because that’s what they sell, manufacturing,” Johnson said.
The US has always led the world in microchip research and design — and that’s what matter most, according to Johnson. For example, Intel is the largest employer in Oregon, where it develops microchip processes and produces them in low volumes for testing before shipping them overseas for manufacturing. Last year, it opened a $3 billion expansion to its global technology development manufacturing facility in Hillsboro, OR.
While the government is incentivising new factories in the US, it doesn't have to do so with research-and-design facilities, which never left the US.
In September 2022, Intel broke ground on a new semiconductor “mega-site” just outside of Columbus, OH that’s expected to accommodate eight chip factories. Intel has said its $20 billion investment in the new mega-site will boost production to meet demand for newer semiconductor technology.
TSMC, which makes 95% of Apple’s iPhone processors, announced late last year that in addition to the company’s first Arizona-based fab — scheduled to begin production of N4 process technology in 2024 — it's begun construction of a second fab set to begin production of 3nm process technology in 2026.
The company is spending roughly $40 billion on the two new plants, representing the largest direct foreign investment in Arizona history and one of the largest such investments in the history of the United States.
TSMC Arizona’s two fabs are expected to create an additional 10,000 high-tech positions, including 4,500 direct TSMC jobs. When complete, TSMC Arizona’s two facilities will manufacture more than 600,000 wafers per year, with an estimated end-product value of more than $40 billion.
“So, first of all, the semiconductor industry never left the US,” Johnson said. “Walk around GlobalFoundries' [facilities] in Malta, NY, or Intel’s fabs in Arizona, New Mexico and Oregon, Micron’s in Utah, Samsung’s in Texas, TI’s in Texas and… and say there’s no manufacturing here,” Johnson said.
“It’s not a question of bringing the whole damn thing back. It’s a security question, really,” he continued. “The world’s an unstable place and having some portion of the critical stuff here in United States makes a lot of sense. Leadership is not volume. Leadership is who controls the technology. Who controls the tech? All the leading chip companies are based in the US.”