TrendForce compiled a brief list of the first customers of TSMC's 2 nm N2 foundry node. Perhaps the biggest customer is Apple, with its upcoming A20 chip powering next-generation iPhones, the M6 line of processors powering its Macs, and the Vision Pro R2 chip powering its next-gen AR wearable, with mass-production expected to commence in 2026. AMD and MediaTek will be among the first customers implementing N2. AMD is developing the CPU Core Dies (CCDs) of its EPYC "Venice" server processors powered by the "Zen 6" microarchitecture; while MediaTek's next flagship mobile SoC, the Dimensity 9600 series, is expected to use this node. Tape-out of the "Zen 6" CCD completed in April 2025, and mass-production is expected in 2026; while Dimensity 9600 just finished tape-out earlier this month, with mass-production expected in late-2026.
www.techpowerup.com/, Sept. 16, 2025 –
This leaves us with NVIDIA, a company that sources some of the largest chips TSMC produces in terms of die-area (the least number of chips per wafer). NVIDIA is preparing its future-generation "Feynman" architecture to succeed its upcoming "Rubin" architecture. White TrendForce points to a 2 nm-class node for "Feynman," with mass-production slated for the second half of 2026, a Commercial Times report TrendForce cites says NVIDIA could opt for the more advanced TSMC A16 node with backside power delivery. A16 would logically be 1.6 nm or 16 Angstrom-class.
Intel is conspicuous in its absence in this list of TSMC 2 nm-class nodes. The company currently uses TSMC 3 nm, specifically the N3B node, for the compute tiles of both its server and client processors. The company is next expected to use its in-house Intel 18A foundry node that implements RibbonFETs and PowerVia technologies for its frontline server and client processors. Intel even put out a first-party announcement to this effect. The company's future chips could implement the more advanced Intel 14A node.