Rumor
Nvidia reportedly redesigning Feynman chips amid TSMC A16 capacity concerns
Monday, March 23, 2026 at 12:24 AM
Nvidia is reportedly planning to redesign its next-generation Feynman chips due to projected capacity constraints on TSMC A16 (1.6nm) process nodes. The company is exploring alternative architectures or process strategies to mitigate potential supply shortages for its future AI infrastructure silicon.
Context
Reports indicate that NVIDIA is planning to redesign its next-generation Feynman architecture, which was recently unveiled at GTC 2026 as the successor to the Vera Rubin platform. The rumored shift is driven by a capacity crunch at TSMC, specifically regarding its A16 (1.6nm) process node. While TSMC officially began volume production of 2nm (N2) chips in late 2025, the highly anticipated A16 node—which features back-side power delivery—is not scheduled for mass production until the second half of 2026.
This supply bottleneck forces NVIDIA to adjust its roadmap for the Feynman generation, which includes the Rosa CPU and LP40 LPU. Because A16 is expected to offer an 8-10% speed improvement and up to 20% power reduction over earlier nodes, competition for initial wafers is "extremely high." To ensure meeting massive demand for agentic AI infrastructure, NVIDIA may be forced to pivot certain designs back to more available 2nm-class variants like N2P or N2X to avoid launch delays.
Sources (9)
NVIDIA GTC 2026: Live Updates on What’s Next in AI | NVIDIA BlogVersion 573.96(Windows) :: NVIDIA Data Center GPU Driver ...TSMC says 1.6nm chips will be in volume production in 2026 • The RegisterWhy Did All 2nm Chips Miss Their Release Schedules in 2025?TSMC's says 1.6nm node to be production ready in late 2026 — roadmap remains on track | Tom's HardwareTSMC plans 1.6nm process for 2026TSMC N2 Process Technology Wiki - SemiWikiNVIDIA Feynman Architecture Introduction: Next-Gen GPUs with TSMC A16 Process - NADDOD Blog
Related Companies
TSMC
2330
Nvidia
NVDA