Intel eyes foundry opening as TSMC’s N3 capacity tightens

TSMC’s N3 process is under mounting pressure as major AI chip designers shift flagship products onto the same node. Nvidia is moving its next-generation Rubin GPUs from 4nm to 3nm, while Google’s TPU Ironwood v7 used N3 and TPU v8 is expected to remain on N3. AI CPUs from Nvidia, Amazon, Microsoft and Arm are also expected to add demand.

TSMC has said N3 capacity could stay tight for several years and announced another expansion in April 2026 after already reaching its planned capacity. Advanced nodes generated 77% of TSMC’s wafer revenue last quarter, while Intel’s external foundry revenue was just $293 million in Q2, compared with TSMC’s total Q2 sales of $40.2 billion.

Intel’s opportunity depends on proving that its 18A and 14A processes can meet performance and yield requirements for major customers. Fab 52 in Arizona has entered full production and is designed to support 40k 18A wafer starts per month, but reported design wins from AMD, NVIDIA, Marvell and others remain unconfirmed. Intel has confirmed Fortinet as an external foundry customer, though that deal uses the older Intel 4 node.