Nvidia Overtakes Apple at TSMC
Nvidia has reportedly overtaken Apple as TSMC's largest customer for its cutting-edge N2 process node. The N2 node is now fully booked until mid-2027, with Nvidia's Vera Rubin AI chips getting top priority. This major supply chain shift could constrain future Apple Silicon launches that rely on the most advanced manufacturing processes.
TSMC's N2 node represents a significant architectural shift from the FinFET transistors used for over a decade to Gate-All-Around (GAA) nanosheet transistors. This GAA structure provides superior control over electrical current, delivering a 10-15% speed increase at the same power or a 25-30% power reduction at the same speed compared to the N3E process. Nvidia's Vera Rubin platform is more than a single chip; it's a "six-chip" ecosystem featuring the Rubin GPU and a Vera CPU with up to 88 ARM cores. The system is built for large-scale AI and introduces HBM4 memory and the sixth-generation NVLink interconnect, which doubles the GPU-to-GPU bandwidth of the prior Blackwell generation to 3.6 TB/s. For nearly a decade, Apple has been the first and largest customer for TSMC's newest process nodes, a strategy that gave its A-series and M-series chips a consistent performance-per-watt advantage. In 2023, Apple accounted for 25% of TSMC's total revenue, while Nvidia was second at 11%. The massive computational and power efficiency demands of AI accelerators are now the primary driver shaping the roadmap for leading-edge process technology, a role previously held by mobile processors. This industry-wide shift in priorities is what enables a data center supplier like Nvidia to command the initial, limited supply of a next-generation node. This supply dynamic could directly influence Apple's M5 silicon roadmap, which is anticipated to use the 2nm process for future Macs and iPads expected in late 2025 and 2026. [cite: 2, 6,
