Growing Pains

The push to redevelop American semiconductor manufacturing has focused heavily on the aptly named Taiwan Semiconductor Manufacturing Company (TSMC), which currently produces the world’s most advanced chips, almost exclusively on the tiny island of Taiwan. First cajoled by the Trump administration and then lured with billions in government support under the CHIPS Act passed in 2022, TSMC has committed to investing $65 billion to build three state-of-the-art facilities in Arizona. Can such industrial policy work? Skeptics jump eagerly on every obstacle and delay to argue no, but the evidence thus far suggests the answer is yes—just not quickly, easily, or cheaply.

Most recently, TSMC updated investors on progress on its April 18 quarterly earnings call. The company confirmed that, after previously postponing its target opening date from 2024 to 2025, it remains on track to begin production in early 2025. Costs in Arizona run higher than in Taiwan, which means customers will be charged more for chips made there. These delays and costs are what analysts and policymakers should have expected. After all, if the U.S. were the best-in-class builder and operator of advanced semiconductor fabs, production would not have been offshored in the first place and no industrial policy would be required to bring it back.

Take the frustrations with the construction process. One recent feature in Rest of World on the “debacle in the desert” blames deep clashes between Taiwanese and American workplace cultures for delays. TSMC has already opened a competing (albeit less advanced) facility in Kumamoto, Japan that it began building a year after it began construction in Arizona. Thus, the story argues, America’s problem must be our incompatible workplace culture uncomfortable with demanding hours and brutal expectations—a clash less profound in Japan.

Differences in work style surely matter, but they ring hollow as a mortal threat to the overall project or an argument against an American industrial policy for semiconductors. For all the challenges that TSMC has faced, it has doubled down not once but twice on its investment since gaining its first experience with the American workforce, expanding its initial $12 billion commitment in 2020 to $40 billion in 2022 after the CHIPS Act passed, and then $65 billion this year after the U.S. government announced it was awarding TSMC $6.6 billion in subsidies, $5 billion in loans, and tax credits for up to 25% of the cost of constructing and equipping its American facilities.

What does TSMC know that the American commentariat does not? Perhaps that the processes and partnerships necessary to do all this well require time to develop. Japan, unlike the U.S., is very good at fostering the support and cooperation critical to these investments. If there is a cultural difference at play, it is Japanese comfort with government-business collaboration, and with the state mechanisms required to effectively attract and promote industry.

Likewise, TSMC CEO C.C. Wei’s confirmation that customers should expect to pay more for chips made in TSMC’s overseas facilities, due to the higher costs of doing business outside Taiwan, should surprise no one. Semiconductor production was offshored for a reason. Tremendous support from foreign governments like Taiwan’s, specifically designed to build leading chip production capacity, left U.S. operations 30–60% more expensive than offshore options. Accordingly, offshore is where production went. The nations pursuing this strategy believed that building their own domestic scale and expertise would allow them to lock-in long-term competitive advantages that would benefit their domestic economies and geopolitical position. They were right.

American economists and policymakers, meanwhile, took the position that such offshoring was beneficial for Americans as well. If Taiwan could produce chips more cheaply—albeit with massive subsidies, at least initially—then all the better for the American consumers who would benefit from lower prices.

Of course, the lower sticker prices disregarded the high costs of all the attendant ills: extreme supply chain fragility resulting in an inability to respond to geopolitical or natural crises; regional economic devastation; the degradation of the industrial commons; and the slowdown in innovation. As Intel’s longtime CEO Andy Grove warned more than a decade ago:

“Transferring manufacturing and a great deal of engineering out of the country, has hindered our ability to bring innovations to scale at home. Without scaling, we don’t just lose jobs—we lose our hold on new technologies. Losing the ability to scale will ultimately damage our capacity to innovate.”

Of course, it will be costly to undo this damage caused by a blind embrace of free trade and disregard for the health of American industry. What’s galling is that those responsible for our predicament are the ones complaining loudly about the price tag for fixing their mistakes.

The crucial point is that these prices are worth paying and are best understood in terms of investment that will pay many future dividends. When one technology is onshored or reshored, ecosystems of suppliers and related technologies spring up around it. In the 1980s, Ronald Reagan responded to the growing dominance of imported Japanese autos with an import quota. Since the restriction applied only to Japanese imports but not to Japanese autos made in America, the result was a massive onshoring of Japanese auto production into the United States. The rest of the production value chain followed. The quota ended after just a few years, but Japanese production continued expanding in the U.S., largely because the efficiencies offered by the ecosystem of suppliers and R&D that had organically grown up around the new production stream. In Reagan’s case, the cost of autos went up 8% in the short term, but his policy prompted three times as much foreign direct investment, and the eventual onshoring of the value chain created over 100,000 American jobs.

This same process is already underway in the semiconductor industry as well.  An underreported moment in TSMC’s earnings call came when one analyst asked Wei about TSMC’s announcement that its third American facility will focus on the most advanced 2nm chips. Would that facility include the ancillary but critical process of advanced packaging, without which the supply chain would remain dependent on Taiwan, or would packaging remain “mostly concentrated in [the] Taiwan region?” Apple, TSMC’s biggest customer, is already partnering with Amkor Technology to provide advanced packaging in Arizona for the chips built by TSMC. On the call, Wei explained that TSMC is working with Amkor to roll out this service to all its Arizona customers. This kind of cooperative effort is a win for the CHIPS Act and a show of force: three major corporations betting that American industrial policy can work, and making their own investments in response. It is just one example of the mass investment that has followed enactment of the CHIPS Act. As AEI senior fellow Chris Miller noted recently in the Financial Times, semiconductor companies and related suppliers have announced over $320 billion in investment over the next decade.

What Grove had to say about manufacturing know-how goes for the task of industrial policy itself. The CHIPS Act’s challenges are the growing pains we should expect from a nation relearning how to foster large-scale, cutting-edge manufacturing after decades of abandoning responsibility for production to the rest of the world. The great British Catholic G.K. Chesterton famously opined that if something is worth doing, it’s worth doing badly. In this case, the growing pains are worth the growth they signify. The pain should serve helpfully to remind us of our national folly in eviscerating both our industrial commons and the state capacity to protect and promote it, and hopefully to ensure that it never happens again.