When Taiwan Semiconductor Manufacturing Company (TSMC) began building massive fabrication plants (“fabs”) in both Phoenix, Arizona, and Kumamoto, Japan, the Japanese investment seemed to offer a seamless path to overcome dangerous vulnerabilities. 

TSMC’s Japanese project finished construction on schedule, buoyed by a harmonious workforce and immediate government subsidies. In contrast, TSMC’s Arizona plant turned into a public relations nightmare, plagued by a year-long delay, ballooning costs, and a high-profile culture clash between Taiwanese management and American unions. 

Yet a year into production, a startling reality has emerged: The “messy” American venture is turning out to be a financial success, while the “perfect” Japanese project is struggling to stay afloat. In 2025, TSMC Arizona posted over $500 million in profit; during the same period, the Japanese operation racked up more than $300 million in losses. Industrial policy can build a factory, but it cannot manufacture demand, and the US factory picked the chips customers prized, while the Japanese factory did not.  

The Japanese and American chip plants came in response to shared concerns that a potential Chinese invasion of Taiwan would cut off the supply of indispensable semiconductors. COVID-19 disruptions then led to a shortfall in chips that caused Japanese giants such as Toyota and US powerhouse Apple to cut production. The lessons were clear: Japan and the US must produce more of their own semiconductors. 

Japan’s strategy represented a marvel of coordination. The Ministry of Economy, Trade and Industry (METI) didn’t just offer subsidies, it orchestrated a public-private consortium. It coordinated an appetizing recipe. Industry giants like Sony and Denso took equity stakes, providing TSMC with instant local expertise and guaranteed customers for automotive and consumer electronics chips. Government subsidies and private investments covered half of TSMC’s construction costs, mitigating their financial risk. Sony even handled land acquisition, utility permits, and specialized workers. Japan’s labor force, accustomed to long hours and corporate loyalty, proved a natural fit and easily adapted to TSMC’s grueling 24/7 operating model. 

But the Japan Inc. embrace came with a bad aftertaste. By anchoring the project to the needs of its domestic partners, Japan bet heavily on “mature-node” chips — the 12-nanometer to 28-nanometer semiconductors used in cars and cameras. Although this initially seemed like a safe, market-driven hedge against COVID-era shortages, the global market had shifted by the time the factory doors opened. A surge in supply from Chinese producers crashed the prices of mature chips, leaving the Japanese fab with a disastrous utilization rate of reportedly around 50%. 

Compared to the Japan Inc. embrace, TSMC ventured alone in the American desert. The US government provided almost no upfront financial support. TSMC obtained funding from the CHIPS and Science Act four years after announcing their Arizona investment. The Taiwanese encountered a deluge of local, state, and federal regulations on land and obtaining permits in Arizona. Advanced semiconductor manufacturing is so novel in Arizona that TSMC claims they had to push new legislation to be created, which cost $35 million. 

TSMC could not transplant its intense managerial style to the American environment. Unlike their Taiwanese counterparts, American workers would not remain on call at all times. Wages are also much higher in the US than in Taiwan. Citing a lack of specialized labor in Arizona, TSMC imported workers from Taiwan, angering local workers. After months of disputes, TSMC and unions struck a nervous deal. 

Then came the upside. Cutting-edge demand from US customers such as Apple and NVIDIA future-proofed Arizona’s fabs. TSMC followed the money toward “advanced nodes” — the tiny, 3-nanometer and 4-nanometer chips that power high-end smartphones and, crucially, artificial intelligence. While the Arizona construction site was a “wasteful, expensive exercise in futility,” according to TSMC founder Morris Chang, the end product was exactly what the world’s most profitable companies were desperate to buy.  

Despite production costs in Arizona being 2.4 times higher than in Taiwan, American technology giants have been willing to pay a premium for American-made chips to secure their supply chains. Today, Arizona’s capacity is so overbooked that customers are reserving space in fabs that haven’t even been built. 

The divergent fates of these two projects offer a pointed lesson for the next decade of “chip wars.” Japan’s experience shows that even the most coordinated industrial policy can become obsolete if it is locked into a specific technology sector. In Japan, TSMC is now playing catch-up, completely reworking their second fab to pivot toward the 3-nanometer advanced chips that Arizona anticipated five years ago. 

The US succeeded almost by accident. Its pioneering technology sector has always used the most advanced semiconductors available. Today, the Arizona fab found itself positioned at the epicenter of the AI boom. Japan’s automotive and electronics sectors do not require the same amount of advanced semiconductors as other industries. This focus on legacy technology caused Japan to lag in the race for next-generation chip manufacturing. 

While policymakers may clamor for domestic chip production in the name of economic security, these efforts are destined to fail without a sustainable market to support them. Kumamoto’s mature nodes may secure Toyota’s and Sony’s supply chain, but TSMC rerouting to produce AI chips demonstrates that supplying the fastest-growing, most advanced markets is crucial. 

The “Arizona Paradox” reveals that the most durable industrial policy is one that remains structurally flexible. As TSMC considers expanding its Arizona footprint to as many as 12 fabs, the lesson for policymakers is clear: you can subsidize the bricks and mortar, and you can streamline the permits, but you cannot legislate demand. The “messy” American fab is working because its customers — the architects of the AI revolution — need it to work. In the high-stakes world of semiconductors, profit follows the market, not the plan. 

Benjamin Echikson is completing a master’s degree in international relations at Johns Hopkins SAIS. He lived and worked for two years in Japan after graduating from Oberlin College and visited Japan on a master’s Capstone project to report on this article.  

Bandwidth is CEPA’s online journal dedicated to advancing transatlantic cooperation on tech policy. All opinions expressed on Bandwidth are those of the author alone and may not represent those of the institutions they represent or the Center for European Policy Analysis. CEPA maintains a strict intellectual independence policy across all its projects and publications.

The post “Messy” US Chip Fab Outperforms Japan’s Textbook Model  appeared first on CEPA.