212. America's semiconductor policy and the AI race with China

Episode Summary

Executive Summary: The episode examines how semiconductors have become central to AI, U.S. economic resilience, and national security. Guest Dan Kim explains why Washington moved from near-indifference to heavy intervention through the CHIPS Act, tariffs, and export controls, highlighting what has worked, what has not, and the ongoing challenge of reducing dependence on Taiwan and China while keeping the U.S. chip ecosystem competitive.

Main Topics: Semiconductors as the foundation of AI and the economy (Priority: 5/5): The hosts frame chips as essential to AI, cloud computing, and broader growth, noting that AI-related stock gains and data center investment have become major economic drivers. Why U.S. policymakers intervened in chips (Priority: 5/5): Dan Kim traces rising concern from China’s Made in China 2025 plan to COVID-era shortages, showing how supply chain fragility and geopolitical risk pushed semiconductors onto the policy agenda. CHIPS Act design and early outcomes (Priority: 5/5): The discussion explains the Act’s mix of grants and tax credits, why industry preferred incentives, and how the program has drawn large announced investments and strategic attention. Site selection and the real requirements for fabs (Priority: 4/5): Kim details that fabs depend on ecosystem factors like workforce, suppliers, water, energy, vibration control, and existing clusters, not just subsidies or cheap labor. Taiwan concentration risk and limits of reshoring (Priority: 4/5): The episode assesses the risks of relying on TSMC and Taiwan for leading-edge chips, while Kim argues CHIPS funding cannot fully offset that exposure. Export controls, tariffs, and unintended consequences (Priority: 5/5): The conversation covers Huawei-era export controls, the October 2022 equipment rules, and the challenge of balancing national security with real costs to U.S. firms and global supply chains. Mature-node chips and the challenge of competing with China (Priority: 4/5): Kim says the U.S. has had limited success attracting mature-node manufacturing because economics are far harder and China’s overcapacity can undercut investment incentives.

Key Arguments: Semiconductors are inseparable from AI because advanced models run on chips and the compute infrastructure they enable. U.S. concern grew when China publicly pursued semiconductor self-sufficiency and when pandemic shortages exposed how fragile supply chains were. The CHIPS Act’s grant-plus-tax-credit structure was designed to lower production costs enough to make U.S. fabs viable again. Government subsidies are only one factor in fab location; supplier ecosystems, workforce, infrastructure, and permitting speed matter just as much or more. The U.S. cannot fully de-risk Taiwan with $39 billion; the scale of TSMC-centered production is too large to replace quickly. CHIPS Act implementation created momentum, attention, and investment, but bureaucracy slowed deployment and frustrated companies. Mature-node production is economically harder to reshore because U.S. greenfield fabs compete against fully depreciated Asian facilities and Chinese overcapacity. Export controls impose genuine revenue and operational costs on U.S. firms and should not be dismissed as painless. Policy should increase supply-chain transparency and phase in restrictions gradually so firms can adjust sourcing and design choices. China’s restrictions can accelerate its own innovation, but U.S. controls also force Chinese firms to improvise and sometimes use more inefficient or less advanced methods.

Data Points: Federal focus on semiconductors in 2015: 2 people - Kim says only two people in the U.S. federal government were focused on semiconductors when he joined the ITC. CHIPS Act grant pool: $39 billion - Amount of the $52 billion CHIPS package allocated for competitive grants. CHIPS Act overall funding: $52 billion - Total CHIPS Act support discussed in the episode. Tax credit rate: 25% - Fully refundable tax credit for eligible capital expenditures under the CHIPS Act. Estimated return on CHIPS commitments: about $400 billion in announced investments - Kim says this followed $39 billion in committed funds, roughly a 10:1 ratio. Updated announced investments: north of $600 billion - Kim says the administration later reported more committed investment. Typical fab cost: around $20 billion+ - Used repeatedly to explain why subsidies and delays matter so much. Suppliers per fab: about 3,000 - Illustrates the complexity of fab ecosystem requirements. U.S. manufacturing share: close to 10% - Kim says U.S. semiconductor manufacturing share declined sharply from earlier dominance. Leading-edge foundry concentration in Taiwan: more than 90% - Kim cites Taiwan’s share of leading-edge foundry manufacturing. Leading-edge companies: 5 companies - Kim identifies the firms able to do leading-edge manufacturing across logic and memory. Mature-node tax credit under new law: 35% - He notes this increased from 25% with the 'big, beautiful bill.' Leading-edge equipment share of wafer cost: 60% to 80% - Kim explains why depreciation dominates leading-edge economics. Potential design-porting cost: hundreds of millions of dollars - Cost to move a chip design from one fab to another. Nexperia chips in a server example: about 15 chips, about $1 total value - Used to illustrate how low-cost mature-node chips can still be system-critical.

Pivotal Quotes: "There is no artificial intelligence without semiconductors." — Dan Kim: Kim explains why chips are foundational to AI systems and models. "Subsidize and they will come, I think it was the theory." — Dan Kim: He summarizes the logic behind the CHIPS Act’s supply-side incentive model. "If you give enough economic signals for their competitors, they could come in and fill it with enough lead time and supply chain managers can deal with it." — Dan Kim: He argues that mature-node supply chains can diversify if policy signals are phased and predictable.

Implications: The episode suggests U.S. chip policy must balance industrial revival, resilience, and security with realistic economics. Future success depends on faster implementation, better demand-side tools, and phased, transparent measures that firms can actually adapt to.