Taiwan Semiconductor Manufacturing Company has solidified its position as the backbone of AI infrastructure by capturing over 90% market share in advanced-node data-center chips. In 2025 the company invested a record $52 billion in capital expenditures—up 30% year-over-year—to expand N3 and N2 capacity across Taiwan and Arizona. Its Arizona Fab 2 began N3 production in Q4, positioning TSMC to supply next-generation AI accelerators for hyperscale customers. By controlling both lithography at 3 nm and below and high-end packaging solutions such as CoWoS, TSMC offers integrated wafer-to-module services that few rivals can match.

Last year the foundry reported revenue of $122.4 billion, a 36% increase versus 2024, marking its first $100 billion year. Gross margin expanded from 56.1% to 59.9%, and operating margin climbed from 45.7% to 50.8%. In Q4 alone, gross margin reached 62.3% and operating margin 54%. These figures reflect both stringent cost controls—operating expenses rose just 2.4%—and significant pricing power in advanced nodes, where wafer prices have risen roughly 20% annually over the past two years without denting order backlogs.

Despite its quasi-monopoly on AI-grade chips and guidance for 20–25% annual revenue growth through 2029, the stock trades at a forward P/E near 25x, a discount of 20–30% to comparable peers in the semiconductor sector. On a PEG basis—adjusting valuation for the 30% earnings growth TSMC projects—its ratio stands below 1.0x versus a sector median above 1.6x. Investors appear to underprice the resilience of net margins above 48% and free cash flow potential exceeding $220 billion in 2026, suggesting a significant re-rating opportunity should the AI super-cycle prove durable.