NVIDIA unveils new AI chip based on TSMC's A16 process at GTC 2026

Summary

At its GTC 2026 conference, Nvidia announced its next-generation AI chip, “Feynman,” which will be the first to use TSMC’s A16 (1.6nm) process technology and is slated for 2028 mass production . Nvidia will have exclusive initial access to this A16 capacity . The Feynman architecture is aimed at physical AI and world models, featuring backside power delivery and offering a significant leap in performance, with a single GPU reaching 50 PFLOPS . Alongside Feynman, Nvidia also detailed its Vera Rubin platform (using a 3nm process) and the new Vera CPU, signaling a broader strategy to provide complete AI data center systems, or “AI Factories” Sina Finance+ 3.

Impact Analysis

This isn’t just a product launch; it’s a strategic move to freeze the market. By announcing the Feynman chip for 2028 on TSMC’s future 1.6nm node , Nvidia is showing a roadmap so far ahead it makes competitors look two generations behind. The market is focused on the headline specs, but the real story is the shift from selling GPUs to selling entire “AI Factories” C114 通信网.

The introduction of the Vera CPU DoNews and the full-rack NVL72 systems money.udn.com shows they’re moving to capture the entire data center value stack, pushing into Intel and AMD’s territory. Locking up initial A16 capacity with TSMC is a massive moat . The execution risk on a 2028 product is obviously high, but the signal is clear: they intend to maintain their monopoly. This makes their $1 trillion revenue forecast feel less like an ambition and more like a concrete plan Sina Finance+ 2.