NVIDIA may redesign its Feynman AI chip platform due to limited 2-nanometer capacity at TSMC. Taiwan’s Economic Daily News reports that TSMC’s advanced manufacturing is fully booked through 2028. High demand from major firms like NVIDIA and Meta drives this production bottleneck.
The Feynman architecture follows the Vera Rubin platform with a scheduled 2028 launch. NVIDIA originally unveiled the next-generation Feynman platform in 2025.
A potential redesign threatens the established timeline and performance targets for future products. TSMC also expects to raise manufacturing prices in response to the sustained high demand.