#AI加速器

Rubin Hasn't Even Ramped Yet, and NVIDIA's Already Sketching Out Feynman: A16 Process Plus the Next Step in Optical Interconnects

NVIDIA's next-generation AI accelerator platform, Feynman, has surfaced ahead of schedule, built on TSMC's A16 process and set to debut CPO optical interconnect technology for the first time. Mass production is expected in 2028 — just as the Rubin generation is barely getting off the ground.