Rumor
Nvidia reportedly secures all initial capacity of TSMC A16 process for Feynman GPUs
Monday, December 22, 2025 at 01:46 AM
Nvidia has reportedly secured the entire initial production capacity of TSMC's A16 backside power delivery process for its upcoming Feynman GPUs, which are expected to succeed the Rubin Ultra series.
Context
Nvidia has reportedly secured the entire initial capacity of TSMC’s upcoming A16 process node for its next-generation Feynman GPU architecture. This 1.6nm-class technology features the Super Power Rail backside power delivery system, which offers substantial improvements in logic density and power efficiency. By monopolizing the node, Nvidia effectively restricts competitors from accessing the industry's most advanced fabrication technology during its early rollout phase.
To balance performance with high-volume availability, the Feynman lineup will likely split across two process nodes. A flagship version will utilize the premium A16 process, while a secondary version will leverage the slightly more mature N2P node. This strategy allows Nvidia to maintain its competitive edge while managing the yield and supply risks associated with pioneering manufacturing techniques. With A16 mass production scheduled for 2026, this move signals Nvidia's intent to sustain its dominance in the high-end AI accelerator market through the end of the decade.
Related Companies
TSMC
2330
Nvidia
NVDA