Jensen Huang pushed back hard on the idea that Google’s TPUs, Meta’s MTIA chips, Anthropic’s multi-gigawatt compute deal, Amazon’s Trainium4, and OpenAI’s custom Broadcom silicon spell the beginning of the end for Nvidia. Speaking on the Dwarkesh Podcast, the Nvidia CEO called Anthropic’s TPU pivot a one-off, defended CUDA’s staying power, and conceded one early mistake—while largely sidestepping the inference era’s uncomfortable cost math. Jensen Huang pushed back hard on the idea that Google’s TPUs, Meta’s MTIA chips, Anthropic’s multi-gigawatt compute deal, Amazon’s Trainium4, and OpenAI’s custom Broadcom silicon spell the beginning of the end for Nvidia. Speaking on the Dwarkesh Podcast, the Nvidia CEO called Anthropic’s TPU pivot a one-off, defended CUDA’s staying power, and conceded one early mistake—while largely sidestepping the inference era’s uncomfortable cost math.
