By Pinelopi Koujianou / Goldberg
Contrary to expectations, the Taiwan Semiconductor Manufacturing Co (TSMC) semiconductor plant in Arizona is reportedly on track to meet its production targets next year. This announcement poses a challenge to the many observers who predicted that the effort to bring chip manufacturing back to the US would fail. What went right this time?
Skepticism surrounding the Arizona plant stemmed from the belief that chip manufacturing benefits heavily from learning-by-doing and dynamic economies of scale, both of which give incumbents a significant cost advantage. That is how TSMC maintains its dominant market position, especially in cutting-edge technologies, producing about 92 percent of the world’s most advanced logic chips at its plants in Taiwan.
It was this high concentration that prompted calls to diversify production in the interest of ensuring supply chain resilience, but the same learning-by-doing imperative is an obstacle for new entrants, casting doubts on projects such as the one in Arizona. Moreover, with anti-immigration sentiment on the rise, concerns about the US’ ability to attract skilled labor added to the pessimism.
Yet the gloomy predictions turned out to be overstated. While learning is crucial in chip manufacturing, new entrants’ ability to compete with incumbents depends more on the nature of the learning. In new research, my colleagues and I find that in semiconductors, learning-by-doing is not as technology-node specific as it is firm specific. Thus, TSMC’s advantage is not necessarily in producing advanced chips more efficiently, but in transferring its knowledge and expertise across different technologies. That means its success could well be replicated in other locations, as long as new plants can build on the Taiwanese parent’s experience, rather than starting from scratch.
Another important factor is the cross-border spillover of knowledge. The same research finds substantial spillovers in the transmission of learning across borders. While the exact mechanisms are unclear, foreign technology transfers — including through foreign direct investment and cross-border recruitment — are likely to play a significant role.
Moreover, since the semiconductor supply chain is structured on the “fabless foundry” model (whereby chip design companies outsource chip fabrication), it, too, facilitates knowledge transfers between nations. Chip design and manufacturing involve close collaboration between buyers and manufacturers around the world, with buyers often a major player in the dissemination of practical knowledge.
These beneficial cross-border knowledge spillovers suggest that government support alone cannot guarantee success in an industry. Taiwan and South Korea owe their dominance in chip manufacturing not only to significant government subsidies, but also to access to foreign advanced technologies.
By contrast, China, despite heavy government support, has not yet reached the frontiers of semiconductor technology. China’s experience shows that while government support can be beneficial, access to foreign technology is crucial. China’s struggles — set against Taiwan’s successes — offer valuable lessons for technologically innovative sectors.
A similar pattern appears in China’s far more successful industrial policy for the automobile sector, where joint ventures between domestic firms and more technologically advanced foreign manufacturers proved instrumental. As with semiconductors, collaboration between firms in different nations stands out as the main driver of technology and improvements in product quality.
Such cross-border learning spillovers have three big implications for current policy. First, the US does indeed stand a strong chance of catching up with Taiwan in semiconductor manufacturing, given the close cooperation between the two nations and the US’ leading position in research and chip design. Second, efforts to slow China’s progress in semiconductors are likely to succeed, considering that US export restrictions have effectively cut China off from foreign advanced technology. Lastly, other nations looking to become major players in the semiconductor industry (such as India) are dependent on US technological leadership. No matter how much financial support these nations provide, they are unlikely to succeed without US technological backing. An industrial policy that might work for the US, the technology leader, would not necessarily work for others.
The key takeaway is that the US remains in the driver’s seat. With its technological leadership and scale, it is poised to meet its semiconductor policy goals — strengthening supply chain resilience and weakening China’s position. Still, one must ask whether these objectives are worthwhile.
While diversifying the supply chain away from a single, geopolitically sensitive location makes sense, it is unclear why chip manufacturing must be brought back to the US, as opposed to other allied nations that might be able to produce at a lower cost. Moreover, the need to slow China’s chipmaking progress remains debatable, except in specific cases where there are legitimate national security concerns.
Historically, the US achieved technological leadership while also lifting up many other nations. The US stayed on top in innovative activities such as research and design, but the gap between it and the rest of the world narrowed. However, the strategy has shifted to one in which the US remains on top by pushing others down.
Yet as US Vice President Kamala Harris argued at the presidential debate: “The true … leader actually understands that strength is not in beating people down, it’s in lifting people up.”
This applies not only to people, but to nations.
Pinelopi Koujianou Goldberg, a former World Bank Group chief economist and editor-in-chief of the American Economic Review, is professor of economics at Yale University.
Copyright: Project Syndicate
What is required for a successful industrial policy
