In 1990, American factories made 37% of the world’s chips. By the time Washington moved to change that, the figure had shrunk to about 12%, according to Computerworld’s December 2022 analysis. The Wall Street Journal had put the same two numbers side by side the year before, citing Boston Consulting Group. The country still leads in chip design and intellectual property, the Journal noted, with Intel, Nvidia and Qualcomm among its homegrown names. Making the chips is another matter.
One company, one island
The Journal’s June 2021 report described Taiwan Semiconductor Manufacturing Co. as the maker of almost all of the world’s most sophisticated chips. Capital Economics estimated its share of those advanced chips at around 92%, with Samsung Electronics producing the rest. TSMC reported $17.6 billion in profits on revenues of about $45.5 billion in the previous year. Taiwanese companies as a group took about 65% of global revenue for outsourced chip manufacturing in the first quarter of 2021, TrendForce estimated, and TSMC alone took 56%.
The Journal drew a regional comparison that readers in the Gulf will recognise: the world’s reliance on Taiwanese silicon resembles its old dependence on Middle Eastern oil, with instability on the island threatening to echo through every industry. Taiwan is also a focal point of tension between the United States and China, which claims the island as its own.
The shortage that followed made the exposure concrete. A US Commerce Department report in January 2022 found that at one point in 2021 there was only a five-day supply of chips worldwide. Commerce Secretary Gina Raimondo said it left auto makers and other users with “no room for error.”
Washington’s answer
The CHIPS and Science Act, signed by President Joseph R. Biden Jr. in August 2022, provides $52.7 billion in manufacturing incentives. Computerworld described its purpose as closing the cost gap with Taiwan, South Korea and China, where governments already subsidise chipmakers. Companies could begin applying for the tax breaks and funds the following year.
Money is already moving. TSMC is building a $12 billion plant in Phoenix, and its 5nm line was set to begin producing in 2024. Bloomberg reported that Apple expected to take about one-third of its future processors from the site. Intel, Micron and Samsung have unveiled American fab plans of their own. Qualcomm and GlobalFoundries said they would invest $4.2 billion to double output at the Malta, New York plant. Intel announced a $20 billion investment in two new US factories in March 2021.
The scale of the contest is hard to overstate. A single fab can cost as much as $20 billion, the Journal reported. TSMC planned to spend $100 billion over three years, nearly a quarter of the entire industry’s capital spending, according to VLSI Research. IC Insights estimated that rivals would need at least $30 billion a year for a minimum of five years to have a reasonable chance of catching TSMC and Samsung.
Limits of the reshoring push
Gaurav Gupta, a Gartner vice president, said the Arizona capacity amounts to “a small fraction” of what TSMC already produces in Taiwan. “US customers will still be reliant on TSMC capacity in Taiwan,” he said. Whether TSMC can match its Taiwan costs and efficiency in the United States, he added, is the open question: “Only time will tell.”
Jack Gold of J. Gold Associates made a similar point about Apple. Chip supply is diversifying, he said, but most finished products are still assembled in Taiwan and increasingly mainland China.
China has made semiconductor independence a pillar of its national strategic plan, the Journal reported, and the European Union has set a goal of producing at least 20% of the world’s next-generation chips in 2030.
