TSMC Expands N3 Capacity to Support Nvidia Ramp

TSMC reported strong Q1 results, raised its 2026 revenue guidance and lifted capital expenditure to support accelerated AI demand, and committed to adding another 3nm-capable fab. Management framed a multiyear AI-driven demand curve while acknowledging near-term capacity tightness and rising cost risks tied to geopolitics.

The company disclosed that the high-performance computing (HPC) segment now accounts for 61% of revenue, reflecting disproportionate demand for AI accelerators and related wafers. Management said N3 contributes a large and growing share of advanced-node sales, with one report citing 34% of advanced-node revenue from 3nm. TSMC plans to expand N3 capacity through multiple channels:

• •adding a new N3 fab in the Tainan GigaFab cluster with first volume in H1 2027
• •bringing a second Arizona N3 fab to volume in H2 2027
• •deploying N3 at a second Japan site (Kumamoto) targeting 2028

The company is also converting N5/N7 tools to support N3 volume and accelerating N2 ramp activities, which are already in high-volume manufacturing with yield improvements. Reported financials include 40.6% year-over-year revenue growth in Q1 and gross margins in the mid-60s range, with analysts flagging meaningful margin expansion driven by the advanced-node mix.

This is an infrastructure-scale reaction to hyperscaler and accelerator vendor demand, most notably from NVIDIA, which is securing large N3 allocations for its next-generation Rubin lineup and HBM4-integrated packages. For practitioners, the implications are concrete: supply-side investments at TSMC are shortening the window of scarcity for advanced AI compute, but the company still expects node-level tightness into 2026-2027. The decision to add fabs and convert capacity underlines the capital intensity required to sustain AI compute growth and the leverage that leading customers exert on foundry capacity planning. The risk profile is asymmetric: higher revenue and margin upside from advanced-node mix versus elevated capital intensity and exposure to geopolitical cost shocks like the Middle East conflict.

Track tool conversions, N3 wafer starts per month at Tainan and Arizona, and Nvidia's capacity commitments; watch gross-margin trajectory as N3 mix increases and as geopolitical cost pressures materialize. Also monitor N2 yield progression and A14 development, which will determine medium-term node economics and whether margin gains from N3 persist.

TSMC's capacity moves affect chip allocation, lead times, and design choices across the AI stack. Teams planning silicon, system design, memory integration, or deployment cost models must assume constrained N3 availability for the next 6-18 months but increasing capacity thereafter. Expect procurement windows to remain strategic and for customers with deep pockets and long-term contracts to retain preferential access.