NVIDIA Selects TSMC’s A16 for Next-Generation Feynman AI Chips

TSMC’s A16 process introduces its Super Power Rail, or SPR, technology, moving power delivery to the back of the wafer to free up front-side routing for signals. Designed for high-performance computing and AI chips, A16 is expected to deliver an 8% to 10% speed gain at the same voltage, or cut power consumption by 15% to 20% at the same speed, compared with N2P. Volume production is scheduled for the second half of 2026.

TSMC has completed validation of A16 with SPR while preserving compatibility with established chip-design tools and intellectual-property ecosystems, easing migration costs for customers. NVIDIA has selected the process for Feynman, its next-generation AI accelerator platform expected in 2028, according to the report. The adoption gives TSMC an early flagship customer as Intel and Samsung continue to contend with design-transition complexity and yield challenges in their competing backside-power technologies.