The United States government is pulling out all the stops to boost domestic semiconductor manufacturing, injecting billions of dollars into the beleaguered sector and flexing all policy muscles available to give it a leg up over competition from Asia.
When the pandemic hit in 2020, firms initially curtailed orders for these micro building blocks needed for smartphones, computers, cars and many other products. Then, as people began working from home, demand soared for information and communication technology – and the chips that power them. A chip shortage ensued, and auto plants had to stop production because they could not obtain chips. This contributed to skyrocketing new and used vehicle prices, a major driver of the painful inflation Americans were feeling.
In a statement earlier this year, Commerce Secretary Gina Raimondo dubbed the semiconductor shortage a “national security” issue because it exposed the dependency of US manufacturing on imports of semiconductors from abroad. Chips also serve critical military applications and are necessary for cybersecurity tools.
The Biden administration and lawmakers rallied in response, passing the CHIPS and Science Act into law in August. The legislation includes $52 billion to strengthen semiconductor manufacturing in the United States. Of this, $39 billion is earmarked for manufacturing incentives, $13.2 billion for research and development and workforce training, and $500 million for international information communications technology security and semiconductor supply chain activities.
Against that backdrop, several prominent companies have announced significant investments in US manufacturing. Taiwan Semiconductor Manufacturing Company (TSMC), a powerhouse in the industry committed at least $12 billion to build a semiconductor fabrication plant in Arizona, with production expected to begin in 2024. At the start of the year, Intel said it planned to build a $20 billion semiconductor manufacturing plant in Ohio, and groundbreaking for the new chip plant took place just last month. And this month, Micron said it would invest up to $100 billion over the next two decades to build a massive semiconductor factory in upstate New York.
In a flurry of tweets earlier this month President Joe Biden pledged: “America is going to lead the way in microchip manufacturing.”
But the US has much catching up to do. US-based fabs, or chip manufacturing plants, currently only account for 12% of the world’s modern semiconductor manufacturing capacity, according to data from the Semiconductor Industry Association trade group. Some 75% of the world’s modern chip manufacturing is now concentrated in East Asia – a majority of that in geopolitically-vulnerable Taiwan. And even with these renewed efforts, the United States does not currently have the same talent and supply chain pipeline as some Asian markets do to support a robust homegrown industry.
To complicate matters, the surge in public and private investments comes at a questionable time, as concerns over the global chip supply shortage have eased. Pandemic-related supply chain blockages are letting up somewhat and a worsening economic outlook has hampered demand.
In an earnings call last week, TSMC CEO C.C. Wei warned it expects the “semiconductor industry will likely decline” in 2023. “TSMC also is not immune,” Wei added, but said it expects “to be more resilient than the overall semiconductor industry.”
Promoting semiconductor manufacturing in the United States now may risk leading to overcapacity and excess supply. And with demand weakening, it isn’t immediately clear if government subsidies will be enough to overcome other obstacles the country faces in developing a competitive semiconductor manufacturing hub.
Understanding the US problems
To understand the latest US efforts, it’s important to be clear on where the country stands – not just in the overall chip industry, but in relation to specific, valuable pockets of it.
“The US is very unlikely to increase its share of global production because even as the US brings online more fab capacity; TSMC, Intel and others are announcing fabs in other places and building them even more quickly,” said Scott Kennedy, a senior adviser at the Center for Strategic and International Studies.
“But I don’t necessarily think that’s really a huge problem,” he added. He noted that measuring manufacturing based on pure output lumps together the lower-end chips and the cutting-edge, higher-end chips that are a more realistic and significant measure of chip manufacturing success. “The US does need to expand chip production for a specific kind of chips, that are directly related to American national security,” he said.
The Biden administration last Friday imposed sweeping new export curbs designed to restrict China’s access to advanced semiconductors made with US equipment, in a move that targets the manufacturing of advanced weapons systems.
While only “about 10% to 14% of chips sold [globally] come from US manufacturing facilities,” according to Columbia Business School professor Dan Wang, the United States does have other strengths. “In terms of design expertise, a lot of that still resides in the U.S.”
Still, the shortcomings are real. “When it comes to foundries, which are the manufacturing side of semiconductors, the U.S. has not really been a major player for many, many years,” said Wang. While it very much used to be, manufacturing began migrating to Asia during the 1980s and ’90s, Wang said. “One of the big reasons for this is that the cost of labor is lower, and it’s just far cheaper to produce at a very massive scale, integrated circuits and chips, in those parts of the world,” Wang added. Morris Chang, the founder of TSMC, said that it costs 50% more to manufacture chips in the U.S. than in Taiwan.
Now, simply having the facilities already set up to produce or expand chip manufacturing gives Asia a big advantage. Wang said he thinks that might be why you see the U.S. “axe-throwing so much money at companies to set up plants in the United States.” It’s not just to respond to demand and become more self-reliant, “but also because you need to get these things up and running very, very quickly, in order to even be in the race at all.”
What it takes to build up a domestic chip industry
Building new chip fabs itself is a costly and time-consuming endeavor. “A modern fab is something like half a million square feet,” said Bob Johnson, an analyst at Gartner, and requires “monstrous clean rooms that have massive air handling capabilities.” He added that these massive buildings require “exceptionally strong foundations.” As he put it, “you cannot have any vibration in the fab because it can wreck the manufacturing process.”
In addition, a single extreme ultraviolet lithography machine, required to map out the circuitry of chips, costs about $150 million, and Reuters reports “a cutting-edge chip plant needs 9-18 of these machines.”
Moreover, the manufacturing of semiconductors requires a range of specialized inputs, including pure chemicals such as fluorinated polyimide, and etching gas, chip etching machines, and more. In places like Taiwan and Fukuoka, Japan, supply chains have developed where the providers of these products are located close to the semiconductor factories. There are also one or two companies that produce vital inputs and that have been trustworthy suppliers to companies in Asia for a long time. This is not yet the case in places like Arizona and Ohio, where plans to build massive chip manufacturing plants are already underway.
You also need a labor force willing and able to do the work.
In the United States, there is both a shortage of new graduates and experienced workers with the technical and engineering knowledge necessary to manufacture semiconductors. Many of those who might have the right experience instead prefer to work in trendier industries, according to Kennedy.
“If we were to today, snap our fingers and have ten new fabs with the world’s leading chips, we probably wouldn’t have enough people to staff them,” Kennedy said. “That’s the biggest bottleneck to the expansion of America’s fab capacity, not capital.”
Intel has tried to establish close relations with Arizona State University to recruit engineers, but it is unclear whether it and other companies building fabs in America will be able to hire enough trained engineers and technicians. If not, even the billions of dollars committed by the private and public sector may not be enough to reshore semiconductor manufacturing.