Taiwan Semiconductor Manufacturing Company (TSMC) is the world's largest dedicated independent semiconductor foundry, holding roughly three-quarters of the global advanced (leading-edge) foundry market and manufacturing the vast majority of the world's most advanced AI accelerators and logic chips, including those designed by Nvidia, AMD, Apple, and Broadcom. Its 3nm and 2nm process nodes and CoWoS advanced-packaging capacity — the binding constraint on AI accelerator supply industry-wide — sit at the center of the global AI buildout, and the company is now the anchor of a multi-decade, multi-country capacity expansion spanning Taiwan, Arizona, and Japan.

• Process technology leadership: Sole high-volume producer at the 3nm/2nm leading edge, with 2nm and A16 capacity targeted to grow at a ~70% CAGR through 2028.
• CoWoS/advanced-packaging bottleneck control: TSMC's CoWoS packaging capacity — not wafer output — is the binding constraint on AI accelerator production industry-wide; capacity is being ramped to 115,000-140,000 wafers/month by end-2026 (>80% CAGR).
• Deep customer lock-in and scale: Multi-year design-in relationships with Nvidia, Apple, AMD, Broadcom, and MediaTek create high switching costs; no credible near-term challenger exists for the customers that matter most at the leading edge.
• Execution track record: Nine consecutive quarters of double-digit profit growth and a history of hitting or beating its own margin guidance through major node transitions.
• Geographic diversification without ceding the technology edge: Arizona and Kumamoto (Japan) expansion reduces single-country concentration for customers while the most advanced nodes remain anchored in Taiwan.

• 2026-08-26 — NVIDIA quantifies its forward capacity book: NVIDIA Corporation disclosed supply and capacity commitments of $279B as of 07/26/2026, covering memory and manufacturing facilities, and guided FY2028 revenue growth to ~70% while calling that figure supply-constrained. TSMC is the counterparty for the logic and advanced-packaging half of that book; the disclosure is the clearest public sizing yet of how much CoWoS and leading-edge wafer capacity NVIDIA has pre-committed, and supports TSMC's raised 2026 capex and above-40% growth guidance.
• 2026-07-25 — Tariff relief: Following the Arizona investment announcement, the Trump administration lowered reciprocal tariffs on Taiwanese goods (from 15% toward roughly 10%), building on a January 2026 US-Taiwan trade deal in which Taiwan committed $250B to U.S. manufacturing in exchange for tariff relief on semiconductors.
• 2026-07-17 — Arizona expansion doubled: Committed an additional $100B to Arizona for at least four more 2nm fabs, lifting total announced U.S. investment to $265B (from $165B) as part of a 12-fab, four-packaging-facility endgame; 2026 capex guidance raised to $60-64B.
• 2026-07-16 — Record Q2 2026 results: Revenue $40.2B (+36% YoY), gross margin 67.7% (above guidance ceiling), net income +77.4% YoY; management raised full-year 2026 revenue growth guidance to above 40%.
• 2026 (ongoing) — Nvidia overtakes Apple as top customer: Nvidia is estimated to generate ~$33B (≈22% of revenue) from TSMC in 2026, surpassing Apple's estimated ~$27B (≈18%), reflecting the AI accelerator ramp; Apple remains the lead customer for the new 2nm (N2) node powering upcoming iPhone (A20), Mac (M6), and Vision Pro chips, alongside AMD and MediaTek.
• 2026 — Kumamoto (Japan) fab upgrade under discussion: TSMC's chairman has discussed upgrading its second Kumamoto fab to 3nm process technology to meet surging AI-driven demand, primarily from US customers such as Nvidia.

• ⚠️ Taiwan/China geopolitical risk: Taiwan's "silicon shield" — the deterrent effect of the island's chip dominance — is widely seen as eroding as China's PLA reportedly targets 2027 for Taiwan-scenario readiness; a blockade/quarantine scenario is considered by some analysts more probable than outright invasion and could disrupt TSMC's Taiwan-based production (still the vast majority of advanced-node capacity) with limited warning.
• ⚠️ Customer concentration: A handful of hyperscaler-adjacent fabless customers (Nvidia, Apple, AMD, Broadcom) account for an outsized share of revenue and of AI-driven growth; a slowdown in AI accelerator demand or a customer in-sourcing/diversification decision would disproportionately hit results.
• ⚠️ Execution risk on overseas buildout: The $265B U.S. expansion and Japan fab upgrades carry substantial capital, labor-cost, and timeline execution risk versus TSMC's core Taiwan operations, where cost structure and yields are proven.
• ⚠️ Valuation and policy risk: Shares trade near a 59% premium to their historical P/E on AI-driven optimism; trade-policy shifts (tariffs, export controls) or an AI capex slowdown among hyperscaler customers are key swing factors.