The hidden supply chain costs of the AI boom

Harvard Business School professor Willy Shih argues that tariffs and trade barriers are prompting companies to reroute or relocate supply chains—not necessarily reshore production—and can raise domestic costs if alternatives are closed off. He says unstable policy makes long-term investment, including semiconductor fabs that need 10–20 years to earn returns, harder to plan, while China retains manufacturing advantages through scale, investment and competition. Shih distinguishes the memory-chip shortage, where AI demand has outstripped capacity, from tariff disruption, which markets have so far partly absorbed through supply-chain adjustments.

Analysis

The investable distinction is between trade friction and physical capacity. Tariffs can redirect assembly and delay cost pass-through; they do not quickly reproduce supplier clusters, skilled labor, or advanced process know-how. That favors firms selling differentiated tools and process capability over a broad “reshoring” basket—but policy volatility can defer the very fab and factory projects those suppliers need. Applied Materials, Lam Research, and ASML therefore have a long-run opportunity, not a clean near-term catalyst. Watch order guidance and customer capex commitments rather than policy announcements.

Apple’s exposure is less a binary China-versus-US call than a concentration problem: high-value components and labor-intensive final assembly cannot be relocated on the same timetable. Tariff escalation or a Taiwan disruption could pressure costs and availability; successful assembly diversification would reduce tail risk but likely require time and execution. Taiwan Semiconductor’s advanced-node concentration leaves similar geopolitical optionality, not a straightforward reshoring beneficiary. Tesla’s Shanghai supplier ecosystem illustrates the productivity advantage of dense clusters; tariffs may shield Western automakers while also raising input costs and weakening competitive pressure. BMW’s globally allocated model production likewise makes tariffs capable of disrupting efficient two-way flows.

Contrarian point: supply-chain resilience can mask, rather than eliminate, cost—front-loading and rerouting postpone repricing. Conversely, a tariff headline alone is not evidence of durable domestic capacity formation. The interview supplies no company-specific financial impact or valuation data, so conviction should stay modest.

Key Decisions for Investors

• No immediate headline-driven position. Track actual tariff implementation, customs enforcement, and company guidance; treat announced factory commitments as unproven until construction, supplier qualification, and output are evident.
• On a 6–18 month horizon, consider a staged long in semiconductor equipment exposure (Applied Materials, Lam Research, or ASML) only if order trends and customer capex confirm sustained investment. Falsify if bookings weaken or major customers delay projects as policy uncertainty persists.
• Keep Apple and Taiwan Semiconductor on a supply-chain risk watchlist rather than shorting on this interview alone. Reassess if tariff costs appear in guidance, assembly diversification stalls, or Taiwan-related risk materially escalates.
• Monitor automotive tariff changes for a relative-value setup: protection may support domestic production, but higher component costs and disrupted model allocation could offset the benefit. Verify company-level guidance and sourcing exposure before positioning in Tesla or BMW.