Executive Summary

The American semiconductor manufacturing industry needs a course correction. The ongoing global chip shortage has shown that U.S. economic and national security depend on secure access to computer chips. But since 1990, the U.S. share of global semiconductor manufacturing capacity has declined while the shares of South Korea, Taiwan, and China have increased. To reverse this trend, Congress passed the CHIPS for America Act, which became law at the start of 2021, authorizing the Department of Commerce to administer tens of billions of dollars in federal grants with the aim of incentivizing chipmakers to increase their manufacturing capacity in the United States

CHIPS Act incentives will help sustain the United States’ advantages as a leader in semiconductor manufacturing. The United States has many features that are appealing to chipmakers, including top talent; the world’s best chip design firms; excellent intellectual property protection; and ample land for developing semiconductor fabrication facilities, known as fabs. U.S. and foreign chipmakers alike have expressed interest in establishing leading-edge chipmaking capacity in the United States with the help of federal incentives. And because of the high economies of scale involved in advanced chipmaking, these initial investments could translate into long-term commitments, helping keep the United States at the leading edge of semiconductor manufacturing for the foreseeable future.

But to be effective, CHIPS Act incentives must be carefully targeted toward specific types of semiconductor capacity. This report therefore assesses (1) the types of chip capacity that are in most urgent need of reshoring from a national security perspective; (2) how much capacity can and should be built for these types of chips with available incentives; and (3) how incentives should be distributed across different types of chips. Note that it is beyond the scope of this report to assess priorities for the $2 billion set aside for mature technology nodes in the U.S. Innovation and Competition Act. This funding should likely go toward the most sensitive needs of the U.S. government, such as specialized analog chips used in military technologies. This report focuses instead on how to spend the remaining $37 billion in manufacturing incentives.

Findings:

• Overdependence on the supply of leading-edge and (to a lesser degree) legacy logic chips manufactured in China and Taiwan threatens U.S. economic and national security.
  • U.S. consumption of logic is worth tens of billions of dollars per year, and an important minority (25 percent) of U.S. logic consumption goes toward more sensitive applications including artificial intelligence, data centers, and the military, as well as automotive applications (including military vehicles).
  • Roughly 85 percent of global leading-edge 5 nanometers (nm) logic manufacturing capacity is located in Taiwan, and roughly 65 percent of global legacy (>16 nm) logic capacity is located in China and Taiwan. In the event of a conflict with China over Taiwan, the United States would likely lose access to all of this capacity.
  • The United States has zero onshore leading-edge (5 nm) logic capacity, and 8 percent of global legacy logic capacity above the 16 nm node.
• Overdependence on the supply of dynamic random access memory (DRAM) chips manufactured in South Korea also poses risks.
  • DRAM chips are also economically vital to the United States, but the country has minimal onshore DRAM manufacturing capacity and none at the leading edge.
  • Half of global DRAM capacity is in South Korea, and most of the remainder (43 percent) is in Taiwan and China. South Korea faces moderate risks of disruptions in manufacturing due to its proximity to North Korea.
• Other types of lower-priority semiconductor devices—flash memory, analog, optoelectronics, sensors, and discretes— pose lower risks of major disruptions.
  • The production of these devices is less concentrated in South Korea, Taiwan, and China; and most of these devices are commoditized and substitutable in the event of interruptions in supply.

Recommendations:

• The United States should reshore enough leading-edge logic capacity to meet 100 percent of U.S. demand through 2027.
  • This brief recommends appropriating at least $23 billion (62 percent) of CHIPS Act incentives for leading-edge logic capacity. This should be sufficient to meet 100 percent of U.S. consumption of advanced logic chips through roughly 2027, and would encourage Intel, Samsung, and TSMC to maintain or establish long-term presences in the United States.
  • Logic chips made by one manufacturer (e.g., Intel) are not direct substitutes for logic chips from another manufacturer (e.g., TSMC) in the event of a shortage. Leading-edge logic incentives should be distributed across Intel, Samsung, and TSMC in proportion to U.S. demand for chips made by each firm. This equates to one fab for Intel, one fab for Samsung, and two fabs for TSMC.
• The United States should reshore one large DRAM fab.
  • Since DRAM exhibits high economies of scale, the minimum cost-effective DRAM footprint is a fab with 100,000 wafers per month (WPM) in capacity. This amounts to roughly 6 percent of current global DRAM capacity, which could address the most sensitive U.S. DRAM demand in the event of a major shortage. This will require $5-10 billion in CHIPS Act funding.
  • DRAM incentives should go to whichever firm can demonstrate the strongest commitment to building future capacity in the United States—perhaps Micron, the only significant DRAM producer headquartered in the United States.
• Any remaining incentives should be used to reshore legacy logic capacity, in coordination with allies.
  • The United States could build between two and five legacy logic fabs with $4-9 billion in incentives. Again, this would not meet full U.S. demand but could be used to cover the most sensitive applications.
  • The United States should also encourage allies— particularly Germany, Japan, and South Korea—to invest in additional legacy capacity, which will help to reduce global dependence on China and Taiwan.
• The United States should engage East Asian allies to assess and coordinate incentives for semiconductor manufacturing.
  • While there appears to be agreement on the role of direct subsidies provided by the Chinese government, further studies and greater transparency are needed to better understand the level of incentives East Asian democracies are providing to their chipmakers.
  • It is unclear whether negotiation to reduce subsidies in China will be successful. However, there may be a greater chance of coordinating with allies, in particular Taiwan and South Korea, to align incentives and encourage supply chain resiliency and diversity.

Alongside funding, Congress should ensure that U.S. chipmakers have access to ample manufacturing talent and onshore advanced packaging capabilities. Congress should also streamline regulatory processes for building new fabs. These aspects are discussed in forthcoming and previous CSET research.