Huang: Nvidia relies 100% on TSMC for flagship AI chips
Nvidia CEO Jensen Huang confirmed the company makes all its flagship GPUs exclusively at TSMC, citing scale, integrated packaging and long-term planning. Samsung still produces Nvidia's Groq3 LPU inference chips on its 4nm process and supplies HBM memory.
- TSMC's foundry revenue is over ten times that of Samsung
- Switching foundries requires redesigning silicon and full re-verification
- Samsung builds Nvidia's Groq3 LPU on 4nm at its Pyeongtaek fab
- TSMC plans roadmaps five years ahead with its customers
Read next
Hardware