TSMC CEO Foresees AI Chip Shortage Lasting Years Despite Capacity Expansion

TSMC’s CEO C.C. Wei said global semiconductor supply will remain insufficient for AI-driven demand for years, even as the company expands capacity in the United States, and that the shortage should continue to support strong revenue growth. At the company’s annual shareholder meeting in Taiwan, Wei reiterated a sales-growth outlook of more than 30% and said it will take a long time before TSMC can fully meet customer needs. He also suggested TSMC is open to higher chip prices to protect profitability but rejected sudden price jumps seen in past memory cycles, emphasizing long-term sustainability. Wei indicated major capital investment will likely stay elevated, with annual capex expected in the roughly $52 billion to $56 billion range, and said he does not currently see signals to cut spending. The AI chip bottleneck is also prompting extended lead times and encouraging some customers to consider alternative suppliers, while competitors such as Intel are positioning themselves as potential “relief valves” for the constrained supply chain. TSMC additionally pointed to profit strength tied to AI and high-performance computing and confirmed plans for employee bonuses to rise by more than 30% this year.
TSMC warned key customers—including NVIDIA and Broadcom—that its most advanced manufacturing nodes are becoming increasingly constrained, contributing to extended lead times and pushing some buyers to look for backup suppliers.
TSMC tied its strong results to AI and high-performance computing, reporting a 35% year-on-year jump in Q4 2025 profit (net income to $16.3 billion) and saying high-performance computing—including AI and 5G—accounted for 55% of total revenue in the quarter.
On the technology front, TSMC said it began mass production of 2nm chips in Q4 2025 and planned to increase output through 2026.
Asked during Q&A when capex might peak and when it could be reduced, C.C. Wei said, "Honestly, I don't know," and added he currently saw no signals that would require cutting capital spending.
Wei said AI demand is broadening beyond data centers into additional end markets—"personal computers, smartphones, automobiles, and Internet of Things (IoT) devices"—as adoption rises across consumer, enterprise, and "sovereign" applications.
Publishers
40
Articles
47
Reach
87