How the United States' CHIPS Act hurts Taiwan

How the United States' CHIPS Act hurts Taiwan

Journalists look at a gallery inside Japan Advanced Semiconductor Manufacturing, a subsidiary of Taiwan Semiconductor Manufacturing (TSMC), at TSMC's new factory in Kikuyo, Kumamoto Prefecture, Japan, on Saturday. (Photo: Bloomberg)
Journalists look at a gallery inside Japan Advanced Semiconductor Manufacturing, a subsidiary of Taiwan Semiconductor Manufacturing (TSMC), at TSMC's new factory in Kikuyo, Kumamoto Prefecture, Japan, on Saturday. (Photo: Bloomberg)

The concentration of advanced semiconductor manufacturing in Taiwan has raised fears in the United States about the vulnerability of this supply chain should China blockade or invade the island. The US CHIPS and Science Act seeks to address that vulnerability with $52 billion (1.86 trillion baht) in subsidies to encourage semiconductor manufacturers to relocate to America. But the legislation, as designed, will fall short of its objective; it may even weaken Taiwan's most important industry, further threatening the island's security.

Today's semiconductor industry is dominated by specialised companies located around the world. TSMC in Taiwan focuses only on contract manufacturing, primarily of high-end chips, whereas other equally important parts of the semiconductor ecosystem include US companies such as AMD, Nvidia, and Qualcomm (which only design chips), the lithography specialist ASML in the Netherlands, Japan's Tokyo Electron (which makes chip-manufacturing equipment), and Britain's Arm (which produces software used to design chips).

All this specialisation offers two main benefits. First, it means that each part of the global supply chain can focus and improve on what it does best, which benefits other parts of the supply chain. Second, global capacity has increased in all segments of the supply chain, which has made the industry more resilient to demand shocks.

The cost of specialisation is that the industry is vulnerable to supply shocks. Although this problem is not isolated to Taiwan -- since all segments of the supply chain are potential choke points -- no other segment faces territorial claims by China. As a result, the US and Japan have offered large subsidies to TSMC to relocate, and TSMC now plans to build new facilities in Kumamoto, Japan, and Phoenix, Arizona.

The facility in Japan will be completed as planned, and many of TSMC's suppliers are also setting up there. But the Phoenix project is already substantially behind schedule, and fewer suppliers have plans to locate there.

TSMC's experience in Camas, Washington (greater Portland) over the last 25 years casts further doubt on the promise of the Phoenix facility. Despite the initial hope that the Portland facility would become TSMC's beachhead in the US market, the company struggled to find the workers it needed to stay competitive. Even after a quarter-century of the same training and the same equipment, production costs are 50% higher than in Taiwan. As a result, TSMC chose not to expand the Portland operation.

The fundamental problem is that while US workers are skilled in chip design, the country lacks workers with the desire or skills necessary for chip manufacturing. Yet specialised skills are critical in this domain. Workers must be meticulous, attentive to detail, and dedicated to consistency, perfection, and timely production. They must have a strong command of the working principles of their equipment -- much of which is highly advanced or customised -- and of data in the field.

TSMC Phoenix will continue to struggle because there simply are too few US workers with the skills necessary for semiconductor manufacturing. Seeking economic security by relocating semiconductor manufacturing to the US is thus an "expensive exercise in futility," as TSMC founder Morris Chang warned in 2022. The $52 billion in the CHIPS Act may seem like a large number, but it will not be enough to create a self-sustaining semiconductor ecosystem in Phoenix.

Industrial policy can work, but only under the right circumstances. TSMC is a testament to that. Taiwan's industrial planners explicitly chose a niche that built on their existing strengths in manufacturing. They did not attempt to replicate Intel, the leading semiconductor company at the time, because too few Taiwanese workers had the necessary design skills. By the same token, Japan's subsidies to lure TSMC are likely to be successful, because Japan already has an ample supply of skilled manufacturing workers.

Like war, industrial policy has many unintended consequences. The availability of free money risks changing TSMC from a company that has relentlessly focused on innovation into one that is more concerned with securing subsidies. The longer it tries to fix its problems in Phoenix, the less attention management will have for other matters. Those problems are so great that they reportedly led to the resignation in December of TSMC Chairman Mark Liu.

The CHIPS Act poses three big risks. For starters, if TSMC does lose its focus on innovation, the biggest losers will be its customers and suppliers, most of which are US firms. The broader AI revolution -- much of it powered by TSMC-made chips -- will grind to a halt. Moreover, TSMC may reduce its investments in capacity in Taiwan, which will make the whole industry less resilient to demand shocks.

Lastly, TSMC may lose its way so much that another company replaces it as the leader in advanced semiconductor manufacturing. Many in Taiwan already view the CHIPS Act as an attempt by the US to grab Taiwan's technology. Taiwanese have taken umbrage at statements by US politicians who claim that Taiwan is a dangerous place to do business, or that the US needs to draw up plans to bomb TSMC's plants and airlift its executives to the US in the event of a Chinese invasion.

TSMC's fall from its dominant position would further reinforce the sense that the US ultimately does not care about Taiwan. Yet if Taiwan's economy and security are undercut, the harm done to America's own national security would outweigh any gain from achieving greater (and costlier) semiconductor capacity in the US. Taiwan's security is ultimately America's security.

Although well-intentioned, the CHIPS Act is poorly designed. Rather than creating a sustainable cluster of semiconductor manufacturing in the US, it is likely to cause long-run damage to TSMC and, ultimately, to Taiwan's economy. It would be far wiser for the US to adopt an approach that protects its own economic security and strengthens Taiwan's at the same time. Committing to defend Taiwan and building capacity in countries such as Japan (where operations are less likely to damage TSMC's business) could be such a strategy. ©2024 Project Syndicate


Chang-Tai Hsieh is Professor of Economics at the University of Chicago. Burn Lin, Dean of the College of Semiconductor Research at National Tsinghua University, is a former vice president of TSMC. Chintay Shih, a professor at National Tsinghua University, is a former president of the Industrial Technology Research Institute.

Do you like the content of this article?
COMMENT (17)