TSMC is massively ramping up its spending to keep pace with an AI sector that has essentially overwhelmed its production capacity. During its Q4 2025 earnings call, the company announced a major increase in capital expenditure (CapEx) for 2026, setting a guidance range between $52 billion and $56 billion. This represents a significant 31% year-over-year jump and lands roughly 17% higher than market consensus.
The foundry’s High-Performance Computing (HPC) segment, which includes AI chips for giants like NVIDIA, has become its primary engine, accounting for 55% to 58% of total revenue. CEO C.C. Wei admitted that while he was initially nervous about the long-term sustainability of AI, he now believes the demand is real and accelerating.
The foundry’s High-Performance Computing (HPC) segment, which includes AI chips for giants like NVIDIA, has become its primary engine, accounting for 55% to 58% of total revenue. CEO C.C. Wei admitted that while he was initially nervous about the long-term sustainability of AI, he now believes the demand is real and accelerating.
Key Strategic Developments
- Arizona Expansion: TSMC is expediting its U.S. operations, with its second Arizona fab (Fab 2) now scheduled to begin production in the second half of 2027—nearly a year ahead of the original 2028 target.
- 2nm Production: Plans are in place to introduce 2nm manufacturing at the Arizona site to support American clients.
- Yield Parity: The company reported that production yields at its new Arizona facilities are already comparable to or even slightly higher (by up to 4%) than its established fabs in Taiwan.
- Advanced Packaging: To address supply bottlenecks, TSMC is aggressively expanding its CoWoS (Chip-on-Wafer-on-Substrate) capacity, aiming for a monthly output of 110,000 to 120,000 units in 2026.
