TSMC Guides 2025 Capex to $52–56 B, Forecasts 25% CAGR Through 2029
TSMC reiterated its AI-led demand for advanced processors by projecting $52–56 billion capex in 2025, driving full utilization of N2/N3/N5 nodes and advanced packaging capacity. Management forecast 25% revenue CAGR through 2029 and gross margins expanding to 63–65% by 2026, underpinned by AI datacenter demand growing at mid-to-high-50s CAGR.
1. Capex Guidance Signals Extended AI Buildout Cycle
Taiwan Semiconductor lifted its 2025 capital expenditure forecast to a supercycle level of $52–56 billion, highlighting plans to expand production through 2028 to meet surging AI demand. Management cited full utilization of N2, N3 and N5 process nodes and aggressive scaling of advanced packaging capacity as key drivers. The updated guidance underpins a projected 25% compound annual revenue growth rate from 2024 through 2029, with AI datacenter sales expected to rise at a mid-to-high 50s percentage CAGR over the same period.
2. Margin Expansion Supports Profitability Outlook
Robust pricing power in advanced logic and an optimized capacity mix are projected to drive gross profit margins to 63–65% in 2026, up from approximately 59% in the fourth quarter of 2025. In Q4 2025, revenue grew 26% year-over-year in U.S. dollar terms, reflecting sustained strength in key customer segments. Continued node transitions to 2-nanometer and 3-nanometer technologies, combined with streamlined manufacturing yields, are expected to further enhance profitability.
3. Attractive Valuation Relative to Growth Profile
Despite commanding leadership in the global AI semiconductor ecosystem and forecasting nearly 30% revenue growth for 2026, the company trades at roughly 24 times forward earnings—below the 30-times multiple typical of large technology peers and only marginally above the broader market’s 22-times multiple. Given superior revenue and profit growth outlooks, this valuation gap presents a compelling opportunity for investors seeking exposure to the foundational ‘picks and shovels’ of the generative AI buildout.