TSMC Reports 35% Q4 Profit Growth, Plans $165 B Arizona Expansion Under $250 B U.S. Pact
TSMC reported Q4 net profit up 35% year-over-year, its eighth straight quarter of growth, and guided 2026 capital expenditures to climb more than 30% at midpoint. The company has earmarked $165 billion for Arizona expansions as part of a $250 billion U.S.-Taiwan semiconductor investment pact.
1. Record Q4 Profit and AI Demand Surge
Taiwan Semiconductor Manufacturing Co. reported a 35% year-over-year increase in fourth-quarter net profit, marking its eighth consecutive quarter of profit growth. The company cited insatiable demand for advanced AI chips as the primary driver, with revenue from its high-performance compute segment rising by more than 40% sequentially. Management noted that utilization at its 5-nanometer and 3-nanometer fabs exceeded 90%, underscoring the tight supply environment for cutting-edge nodes.
2. Accelerated U.S. Expansion Under New Trade Deal
Under the recently announced U.S.-Taiwan trade agreement, TSMC committed to boost its investment in American manufacturing by an additional $100 billion, bringing its total U.S. investment to $265 billion. The deal includes a reduction of U.S. tariffs on Taiwanese imports from 20% to 15% and full elimination on select goods. TSMC has already purchased two parcels of land in Arizona and plans to construct a multi-fab ‘gigafab’ cluster, which is expected to add up to 200,000 wafer starts per month by 2028.
3. Capital Expenditure Guidance Raised for 2026
Following its blowout quarter, TSMC updated its capital expenditure forecast for 2026, guiding for a mid-point increase of more than 30% compared to 2025’s outlay of approximately $36 billion. CFO Wendell Huang emphasized that elevated spending will support capacity expansion in both Taiwan and the U.S., particularly for AI-optimized process technologies. This elevated capex plan underscores the company’s strong conviction in the long-term AI megatrend and its intention to narrow any supply-demand gap in the most advanced nodes.